COST/CAPACITY

Prices and limits, with date.

The published cost is only comparable when we know the unit, modality and conditions.

Published data

USD per million tokens when the source uses that unit.

OpenAI · Text and reasoning

GPT‑6 Astra

Standard input
10.00 USD / 1 M tokens
Standard output
50.00 USD / 1 M tokens
Cached input
1.00 USD / 1 M tokens
Batch input
5.00 USD / 1 M tokens
Batch output
25.00 USD / 1 M tokens
Context
1,050,000 tokens
maximum output
128,000 tokens

Requests with more than 272,000 input tokens apply price multipliers; tools may add per-call charges.

Open profile and sources →
Anthropic · Text and reasoning

Claude Opus 5

Standard input
5.00 USD / 1 M tokens
Standard output
25.00 USD / 1 M tokens
Cached input
0.50 USD / 1 M tokens
Batch input
2.50 USD / 1 M tokens
Batch output
12.50 USD / 1 M tokens
Context
1,000,000 tokens
maximum output
128,000 tokens

Thinking is billed as output. Fast mode costs twice the base rate.

Open profile and sources →
Google · Text and reasoning

Gemini 3.8 Flash

Standard input
0.75 USD / 1M tokens
Standard output
3.75 USD / 1M tokens
Cached input
0.075 USD / 1 M tokens
Batch input
0.375 USD / 1 M tokens
Batch output
1.875 USD / 1 M tokens
Context
1,048,576 entry tokens
maximum output
65,536 output tokens

Introductory pricing until December 31, 2026; from 2027, standard rates double.

Open profile and sources →
OpenAI · Text and reasoning

GPT‑5.6 Sol

Standard input
4,00 USD / 1 M tokens
Standard output
20,00 USD / 1 M tokens
Cached input
0,40 USD / 1 M tokens
Batch input
2,00 USD / 1 M tokens
Batch output
10.00 USD / 1 M tokens
Context
1,050,000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
OpenAI · Text and reasoning

GPT‑5.6 Terra

Standard input
2,00 USD / 1 M tokens
Standard output
12,00 USD / 1 M tokens
Cached input
0,20 USD / 1 M tokens
Batch input
1.00 USD / 1 M tokens
Batch output
6,00 USD / 1 M tokens
Context
1,050,000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
OpenAI · Text and reasoning

GPT‑5.6 Luna

Standard input
0,20 USD / 1 M tokens
Standard output
1,20 USD / 1 M tokens
Cached input
0,02 USD / 1 M tokens
Batch input
0,10 USD / 1 M tokens
Batch output
0,60 USD / 1 M tokens
Context
1,050,000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Anthropic · Text and reasoning

Claude Fable 5.1

Standard input
10.00 USD / 1 M tokens
Standard output
50.00 USD / 1 M tokens
Cached input
1.00 USD / 1 M tokens
Batch input
5.00 USD / 1 M tokens
Batch output
25.00 USD / 1 M tokens
Context
1,000,000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Anthropic · Text and reasoning

Claude Sonnet 5

Standard input
2,00 USD / 1 M tokens
Standard output
10.00 USD / 1 M tokens
Cached input
0,20 USD / 1 M tokens
Batch input
1.00 USD / 1 M tokens
Batch output
5.00 USD / 1 M tokens
Context
1,000,000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Anthropic · Text and reasoning

Claude Haiku 4.5

Standard input
1.00 USD / 1 M tokens
Standard output
5.00 USD / 1 M tokens
Cached input
0,10 USD / 1 M tokens
Batch input
0.50 USD / 1 M tokens
Batch output
2.50 USD / 1 M tokens
Context
200.000 tokens
maximum output
64.000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Google · Text and reasoning

Gemini 3.1 Pro

Standard input
2,00 USD / 1 M tokens (<200k)
Standard output
12,00 USD / 1 M tokens (<200k)
Cached input
0,20 USD / 1 M tokens
Batch input
1.00 USD / 1 M tokens
Batch output
6,00 USD / 1 M tokens
Context
1.048.576 tokens
maximum output
65.536 tokens

A higher pricing tier applies above 200,000 tokens.

Open profile and sources →
xAI · Text and reasoning

Grok 4.6

Standard input
2,00 USD / 1 M tokens
Standard output
6,00 USD / 1 M tokens
Cached input
0.50 USD / 1 M tokens
Batch input
Not published
Batch output
Not published
Context
500.000 tokens
maximum output
No published text limit

Input and output pricing doubles above 200,000 tokens.

Open profile and sources →
DeepSeek · Text and reasoning

DeepSeek V4.1 Flash

Standard input
0,30 USD / 1 M tokens (peak)
Standard output
1,20 USD / 1 M tokens (peak)
Cached input
0,006 USD / 1 M tokens (peak)
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
384.000 tokens

Off-peak rates are half price; schedules are published in UTC.

Open profile and sources →
Mistral AI · Text and reasoning

Mistral Medium 3.5

Standard input
1,50 USD / 1 M tokens
Standard output
7,50 USD / 1 M tokens
Cached input
0,15 USD / 1 M tokens
Batch input
Not published
Batch output
Not published
Context
256.000 tokens
maximum output
Not published

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Meta · Text and reasoning

Llama 4 Maverick

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
Not published

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Cohere · Text and reasoning

Command A+

Standard input
Free up to the quota limit
Standard output
Free up to the quota limit
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128,000 tokens
maximum output
64.000 tokens

Production use is offered through Cohere Model Vault and enterprise agreements.

Open profile and sources →
Amazon · Text and reasoning

Amazon Nova 2 Lite

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
65.536 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Alibaba Qwen · Text and reasoning

Qwen3‑Max Thinking

Standard input
1,20 USD / 1 M tokens (≤32k)
Standard output
6,00 USD / 1 M tokens (≤32k)
Cached input
0,24 USD / 1 M tokens
Batch input
0,60 USD / 1 M tokens (≤32k)
Batch output
3,00 USD / 1 M tokens (≤32k)
Context
262.000 tokens
maximum output
65.000 tokens

Higher pricing tiers apply above 32,000 tokens.

Open profile and sources →
Z.AI · Text and reasoning

GLM‑5.1

Standard input
1,40 USD / 1 M tokens
Standard output
4,40 USD / 1 M tokens
Cached input
0,26 USD / 1 M tokens
Batch input
Not published
Batch output
Not published
Context
200.000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
MiniMax · Text and reasoning

MiniMax M2.7

Standard input
0,30 USD / 1 M tokens
Standard output
1,20 USD / 1 M tokens
Cached input
0,06 USD / 1 M tokens
Batch input
Not published
Batch output
Not published
Context
204.800 tokens
maximum output
Not published

Consult the primary source before budgeting for a deployment.

Open profile and sources →
OpenAI · Text and reasoning

GPT‑5.5

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
400,000 tokens
maximum output
128,000 tokens

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
OpenAI · Code

GPT‑4.1

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,047,576 tokens
maximum output
32,768 tokens

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
OpenAI · Text and reasoning

gpt‑oss‑120b

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
131,072 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
OpenAI · Image

GPT‑Image‑2.5 Sunburst

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Prompt and reference images
maximum output
Image

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
OpenAI · Video

Sora 2 Pro

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not applicable or not published
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
OpenAI · Voice and audio

GPT‑Live 1

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Real-time session
maximum output
Streaming audio and text

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
OpenAI · Voice and audio

GPT‑Transcribe

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Audio via file or streaming
maximum output
Transcription

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
OpenAI · Embeddings

text‑embedding‑3‑large

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
8,191 tokens
maximum output
Up to 3,072 dimensions

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Anthropic · Text and reasoning

Claude Opus 4.5

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
200.000 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Anthropic · Code

Claude Sonnet 4.5

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
200.000 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Google · Text and reasoning

Gemini 3.7 Flash

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1.048.576 tokens
maximum output
65.536 tokens

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Google · Image

Nano Banana 2

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Multimodal prompt
maximum output
Image

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Google · Video

Veo 3.1

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Prompts and visual references
maximum output
Video clips with audio

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Google · Music

Lyria 3.5

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Creative prompt and optional image
maximum output
Songs up to 3 minutes

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Google · Embeddings

Gemini Embedding 2

Standard input
0.20 USD / 1M text tokens
Standard output
No separate output charge
Cached input
Not published
Batch input
0.10 USD / 1M text tokens in Batch
Batch output
Not published
Context
8,192 input tokens
maximum output
Vectors with 128–3,072 dimensions

Gemini API charges per million input tokens: text USD 0.20, images USD 0.45, audio USD 6.50 and video USD 12. Batch reduces these rates to USD 0.10, 0.225, 3.25 and 6, respectively. Numeric comparison values represent text only. The free tier has its own limits.

Open profile and sources →
Google · Robotics

Gemini Robotics ER 2

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Visual and instruction input
maximum output
Plans and tool calls

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Meta · Multimodal

Llama 4 Scout

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
10,000,000 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Meta · Vision and documents

Segment Anything Model 3

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Image or video
maximum output
Masks and tracking

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
DeepSeek · Text and reasoning

DeepSeek V3.2

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128,000 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
DeepSeek · Text and reasoning

DeepSeek R1

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
64,000 tokens in the original API
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Mistral AI · Text and reasoning

Mistral Small 4

Standard input
0,15 USD / 1 M tokens
Standard output
0,60 USD / 1 M tokens
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
256.000 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Mistral AI · Voice and audio

Voxtral TTS

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Text and voice reference
maximum output
Streaming audio

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Mistral AI · Vision and documents

Mistral OCR 4.1

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Document or image
maximum output
Text and structure

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Alibaba Qwen · Text and reasoning

Qwen3 235B A22B

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128,000 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Alibaba Qwen · Code

Qwen3‑Coder 480B A35B

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
256,000 native tokens · up to 1M extended
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Microsoft · Multimodal

Phi‑4 Multimodal

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128,000 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Cohere · Embeddings

Cohere Embed 4

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128,000 tokens
maximum output
256, 512, 1,024 or 1,536 dimensions

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Cohere · Text and reasoning

Command R 08‑2024

Standard input
0,15 USD / 1 M tokens
Standard output
0,60 USD / 1 M tokens
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128,000 tokens
maximum output
4,000 tokens

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Stability AI · Image

Stable Diffusion 3.5 Large

Standard input
6.5 credits per generation
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Prompt of up to 10,000 characters
maximum output
1 MP image in API

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
Stability AI · Music

Stable Audio 3.0

Standard input
26 credits per generation
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Prompt and reference audio
maximum output
Up to 6 minutes of audio

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
ElevenLabs · Voice and audio

Eleven v3

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
5,000 characters per request
maximum output
Audio

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
ElevenLabs · Voice and audio

Scribe v2

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Audio or video file
maximum output
Structured transcription

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
NVIDIA · Text and reasoning

NVIDIA Nemotron 3 Ultra

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
Not published

Consult the primary source: the billing unit depends on the model type and access channel.

Open profile and sources →
webAI · Text and reasoning

TwIL-8B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
131,072 tokens
maximum output
Not published

Weights are available under a non-commercial license. webAI publishes no API price for this checkpoint. Self-hosting requires compute resources.

Open profile and sources →
Tencent · Inference acceleration

DSpine for Qwen3.8-27B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
8,192-token sliding window in the draft
maximum output
Blocks of 16 candidate tokens; target-model verification

The release provides Apache 2.0 weights, not a commercial per-token API. Running them also requires the target model and compute resources. Unknown rates remain unpublished; they are not presented as free, and throughput is not used to calculate savings.

Open profile and sources →
OpenAI · Code

GPT-6 Cyber

Standard input
20 USD / 1M input tokens
Standard output
100 USD / 1M output tokens
Cached input
2 USD / 1M cached input tokens
Batch input
Not published
Batch output
Not published
Context
1,050,000 context tokens; up to 922,000 input
maximum output
128,000 output tokens

Standard short-context rates per million tokens: USD 20 input, USD 2 cached input and USD 100 output. Cache writes cost USD 25 for 30 minutes or USD 50 for 12 hours. These figures do not budget every context length or include possible tool charges.

Open profile and sources →
Google · Video

Gemini Omni 1.2 Flash

Standard input
1.50 USD / 1M input tokens
Standard output
9 USD text · 12 USD video / 1M tokens
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
131,072 input tokens
maximum output
3–15 s video; up to 65,536 output tokens on Gemini API

Paid Gemini API rates are USD 1.50 input, USD 9 text output and USD 12 video output per million tokens. At 720p this is about USD 0.07 per generated second before input charges. Resolution and usage affect cost; text and video have separate output rates.

Open profile and sources →
Thinking Machines Lab · Multimodal

Inkling 2 Small

Standard input
0.30 USD / 1M input tokens
Standard output
1 USD / 1M output tokens
Cached input
0.03 USD / 1M cached input tokens
Batch input
Not published
Batch output
Not published
Context
Up to 1M in the model card; Tinker offers 64K and 256K
maximum output
Not published

Tinker charges USD 0.30 input, USD 0.03 cached input and USD 1 output per million tokens. Training is separate at USD 1.50 with 64K or USD 3 with 256K. Serverless inference has a separate identifier and beta access; confirm the mode before budgeting.

Open profile and sources →
Baidu · Text and reasoning

ERNIE 5.5 Flash Base

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128K during pretraining; 1M for tuning and inference
maximum output
Not published

No API price has been verified for the base checkpoint. Local compute and license compliance involve separate costs. Check the official commercial-use requirements; this is not an Apache license or a guarantee of free inference.

Open profile and sources →
Ai2 · Robotics

Molmo2-4B-tb

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
One 640 × 360 front image and a text goal
maximum output
Not published

No hosted commercial API has been verified for this checkpoint. Downloadable weights do not cover storage, accelerators, memory or operating costs.

Open profile and sources →
Ai2 · Robotics

ManiFlow-tb

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
224 × 224 front and gripper images; 9-dimensional state
maximum output
25 actions executed from a horizon of 50

No hosted commercial API has been verified for this checkpoint. Downloadable weights do not cover storage, accelerators, memory or operating costs.

Open profile and sources →
ByteDance Seed · Interpretability

Cross-Layer TopK CLT

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
36 layers and 32,768 features per layer
maximum output
360 active features per token across layers

No hosted commercial API has been verified for this checkpoint. Downloadable weights do not cover storage, accelerators, memory or operating costs.

Open profile and sources →
ByteDance Seed · Interpretability

Qwen3-30B-A3B MoE-Output SAE

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Qwen3-30B-A3B layers 4, 30 and 46
maximum output
100 active features per token per autoencoder

No hosted commercial API has been verified for this checkpoint. Downloadable weights do not cover storage, accelerators, memory or operating costs.

Open profile and sources →
OpenAI · Image

GPT Image 2.6 Sunburst

Standard input
5 USD text · 8 USD image / 1M tokens
Standard output
30 USD / 1M image tokens
Cached input
1.25 USD text · 2 USD image / 1M tokens
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Image; SVG and GIF through documented workflows

Prices per million tokens are 5 USD for input text, 1.25 USD for cached text, 8 USD for input images, 2 USD for cached images and 30 USD for output images. These are not per-image prices. Size, quality and usage change the total. Text and image tokens have different rates and must be budgeted separately.

Open profile and sources →
OpenAI · Image

GPT Image 2.6 Flare

Standard input
5 USD text · 8 USD image / 1M tokens
Standard output
30 USD / 1M image tokens
Cached input
1.25 USD text · 2 USD image / 1M tokens
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Image; SVG and GIF through documented workflows

Prices per million tokens are 5 USD for input text, 1.25 USD for cached text, 8 USD for input images, 2 USD for cached images and 30 USD for output images. These are not per-image prices. Size, quality and usage change the total. Text and image tokens have different rates and must be budgeted separately.

Open profile and sources →
Tencent · Multimodal

Tencent Hy4

Standard input
6 CNY / 1M tokens before discount
Standard output
18 CNY / 1M tokens before discount
Cached input
0.3 CNY / 1M tokens before discount
Batch input
Not published
Batch output
Not published
Context
1M tokens; maximum 960K input on TokenHub
maximum output
64K tokens on TokenHub

TokenHub lists base rates of 6 CNY input, 0.3 CNY cached input and 18 CNY output per million tokens. A 50% discount applies until December 31, 2026 at 23:59:59 Beijing time. Prices are not converted to USD using an assumed exchange rate. OpenRouter rates are independent.

Open profile and sources →
Ricursive Intelligence · Code

Ripple-65B Preview

Standard input
0.20 USD / 1M tokens on OpenRouter
Standard output
0.60 USD / 1M tokens on OpenRouter
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Up to 1M tokens according to the model card
maximum output
128K tokens on OpenRouter

OpenRouter lists 0.20 USD input and 0.60 USD output per million tokens for its channel. These are hosting rates, independent of deploying the Apache 2.0 weights. Local cost depends on hardware and load. No Batch discount has been verified for this channel.

Open profile and sources →
Ricursive Intelligence · Text and reasoning

Ripple-65B Base Preview

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Pretraining at 8K; useful long-context capability unverified
maximum output
Not published

The weights require your own infrastructure. Cost depends on memory, precision, sequence length and load. Review the license and runtime code before deployment. A published performance figure does not guarantee the same cost or speed on another machine.

Open profile and sources →
Ricursive Intelligence · Inference acceleration

Ripple DFlash2

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
131,072 configured positions
maximum output
Blocks of 16 candidate tokens

The weights require your own infrastructure. Cost depends on memory, precision, sequence length and load. Review the license and runtime code before deployment. A published performance figure does not guarantee the same cost or speed on another machine.

Open profile and sources →
Sarvam AI · Classification and decision

Sarvam Prism

Standard input
4 INR / 1M tokens in the API
Standard output
0 INR for output in the API
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
API: 32,768 tokens per question; weights: 131,072
maximum output
Probability per option, without text generation

Sarvam lists 4 INR per million input tokens and 0 INR for output tokens. This is decision API billing, not a price for generated text. Local deployment adds infrastructure and operations. No USD conversion is invented and GPU resources are not assumed to be free.

Open profile and sources →
NVIDIA · Vision and geometry

MetricTrail 1.0

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Time-ordered inventory of at least 16 images
maximum output
NPZ files and JSON records

The weights require your own infrastructure. Cost depends on memory, precision, sequence length and load. Review the license and runtime code before deployment. A published performance figure does not guarantee the same cost or speed on another machine.

Open profile and sources →
Moonshot AI · Multimodal

Kimi K3.1 Pro

Standard input
See the API channel’s official pricing
Standard output
See the API channel’s official pricing
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1.048.576 tokens
maximum output
Not published

This profile covers the downloadable checkpoint and retains a link to the official terms. API, Batch and cache rates must be checked for the channel used. Deploying 2.8 trillion parameters depends on runtime, quantization and distribution; downloadable weights do not mean free inference.

Open profile and sources →
ElevenLabs · Audio and music

ElevenLabs SFX v3

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Between 0.5 and 30 seconds per request

Billing depends on the ElevenLabs plan, credits and format. A single USD rate per request or equivalence with text-to-speech pricing has not been verified. Check generated-audio cost and the chosen plan’s commercial usage terms.

Open profile and sources →
Meituan · Multimodal

LongCat-2.5

Standard input
0.30 USD / 1M tokens
Standard output
1.20 USD / 1M tokens
Cached input
0.006 USD / 1M tokens
Batch input
Not published
Batch output
Not published
Context
1.048.576 tokens
maximum output
262,144 tokens

Official promotional USD rates per million tokens are 0.30 for uncached input, 0.006 for cached input and 1.20 for output. These are temporary discounts with no published end date; the platform and invoice take precedence. CNY prices are also published and are not converted using an assumed exchange rate. No Batch discount has been verified.

Open profile and sources →
Microsoft · Science and simulation

Graphormer Redox · Oxidation

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Molecular graphs; limit not published
maximum output
Estimated redox potential

The repository declares MIT. No commercial API or per-query pricing has been verified. The weights require your own infrastructure and code review before loading PyTorch files.

Open profile and sources →
Microsoft · Science and simulation

Graphormer Redox · Reduction

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Molecular graphs; limit not published
maximum output
Estimated redox potential

The repository declares MIT. No commercial API or per-query pricing has been verified. The weights require your own infrastructure and code review before loading PyTorch files.

Open profile and sources →
NVIDIA · Science and simulation

STRATA v1

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
149 input channels
maximum output
148 channels · 10-minute step

Weights use OpenMDW 1.1 and code uses Apache 2.0. Some auxiliary fields require CC BY 4.0 attribution. No API pricing is verified; costs depend on GPUs, data and infrastructure. We did not download weights or run the model for this profile.

Open profile and sources →
NVIDIA · 3D and motion

SOMA-X v0.3.5

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Pose and identity parameters
maximum output
Body and hand geometry

The project declares Apache 2.0, with separate terms for external resources such as MANO or SMPL. No commercial API or per-token pricing is verified. Budget for infrastructure and review each backend’s licenses.

Open profile and sources →
Z.AI · Text and reasoning

GLM-5.3-FlashX-Legal

Standard input
0.37 USD / 1M tokens
Standard output
1.25 USD / 1M tokens
Cached input
0.075 USD / 1M tokens
Batch input
Not published
Batch output
Not published
Context
1M tokens
maximum output
128K tokens

Official USD rates per million tokens: 0.37 input, 0.075 cached input and 1.25 output. Tools may add charges; AutoClaw and other intermediaries have their own terms. No Batch discount or downloadable weights have been verified for this variant.

Open profile and sources →
Microsoft · Vision and documents

Mage-VL-Cowork

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 tokens in reference training and inference
maximum output
Not published

Apache 2.0 weights for self-hosting. No commercial API or price has been verified. Task-synthesis costs in the report are not inference rates; budget for GPU, memory and the execution environment.

Open profile and sources →
OpenAI · Text and reasoning

GPT-6 Rosalind Research

Standard input
15.00 USD / 1M tokens
Standard output
75.00 USD / 1M tokens
Cached input
1.50 USD / 1M tokens
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Not published

OpenAI publishes standard rates per million tokens of USD 15 input, USD 75 output and USD 1.50 cached input. Cache writes cost USD 18.75 for 30 minutes and USD 37.50 for 12 hours. Confirm context, contract and surcharges before budgeting.

Open profile and sources →
Mistral AI · Text and reasoning

Mistral Large 4

Standard input
0.70 USD / 1M tokens
Standard output
2.10 USD / 1M tokens
Cached input
0.07 USD / 1M tokens
Batch input
Not published
Batch output
Not published
Context
1M tokens
maximum output
Not published

Standard API rates in USD per million tokens, verified on October 30. Self-hosting has additional costs. The weights license modifies MIT and requires a separate commercial license to offer Model-as-a-Service.

Open profile and sources →
inclusionAI · Text and reasoning

Ling 3.0 Flash Sante FP8

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
256K tokens
maximum output
Not published

MIT weights for self-hosting, with no verified hosted pricing for this checkpoint. Memory and performance depend on FP8 support and the server; numerical equivalence with BF16 is not assumed.

Open profile and sources →
OpenAI · Text and reasoning

GPT-6 Luna · ChatGPT

Standard input
Depends on the ChatGPT plan
Standard output
Depends on the ChatGPT plan
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Not published

No per-token pricing has been verified for this ChatGPT update. Numerical costs remain empty; subscriptions and API billing are different concepts.

Open profile and sources →
Google · Video

Genie 4

Standard input
No verified public pricing
Standard output
No verified public pricing
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Interactive scenes and controls; no published token limit
maximum output
Real-time video and audio; sessions lasting a few minutes

No prices, quotas or downloadable weights have been published. An Ultra subscription does not establish current access to Genie 4. Preview terms depend on the access program.

Open profile and sources →
NVIDIA · Voice and audio

Nemotron 3.5 Transcribe 30B A4B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
65,536 tokens; recordings up to one hour
maximum output
Not published

BF16, FP8 and NVFP4 are publicly available under OpenMDW 1.1, with access verified on October 30. No commercial per-token pricing has been verified. Budget for GPU, memory and operation; each precision needs its own hardware and validation.

Open profile and sources →
Liquid AI · Text and reasoning

LFM3-1.7B-Base

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
256K tokens
maximum output
Not published

This profile covers downloadable weights. No API rate has been verified for this checkpoint. The LFM license has its own commercial terms; it is not MIT or Apache 2.0. Budget for memory, computing and operations too.

Open profile and sources →
Liquid AI · Text and reasoning

LFM3-1.7B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
256K tokens
maximum output
Not published

This profile covers downloadable weights. No API rate has been verified for this checkpoint. The LFM license has its own commercial terms; it is not MIT or Apache 2.0. Budget for memory, computing and operations too.

Open profile and sources →
Liquid AI · Vision and documents

LFM3-VL-2B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128K tokens
maximum output
Not published

This profile covers downloadable weights. No API rate has been verified for this checkpoint. The LFM license has its own commercial terms; it is not MIT or Apache 2.0. Budget for memory, computing and operations too.

Open profile and sources →
Liquid AI · Voice and audio

LFM3-Audio-2B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32K tokens
maximum output
Not published

This profile covers downloadable weights. No API rate has been verified for this checkpoint. The LFM license has its own commercial terms; it is not MIT or Apache 2.0. Budget for memory, computing and operations too.

Open profile and sources →
Liquid AI · Multimodal

LFM3-Omni-2B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32K tokens
maximum output
Not published

This profile covers downloadable weights. No API rate has been verified for this checkpoint. The LFM license has its own commercial terms; it is not MIT or Apache 2.0. Budget for memory, computing and operations too.

Open profile and sources →
Liquid AI · Text classification

d1-2B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128K tokens
maximum output
Not published

This profile covers downloadable weights. No API rate has been verified for this checkpoint. The LFM license has its own commercial terms; it is not MIT or Apache 2.0. Budget for memory, computing and operations too.

Open profile and sources →
Liquid AI · Text classification

d1-Omni-2B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128K tokens
maximum output
Not published

This profile covers downloadable weights. No API rate has been verified for this checkpoint. The LFM license has its own commercial terms; it is not MIT or Apache 2.0. Budget for memory, computing and operations too.

Open profile and sources →
Alibaba Qwen · Voice and audio

Qwen Audio 3.1 Realtime Flash

Standard input
0.23 USD / 1M text tokens
Standard output
0.70 USD / 1M text tokens
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
262,144 tokens
maximum output
16,384 tokens

Singapore, USD per million tokens: text input 0.23 and output 0.70; audio 0.93 and 1.87. When a response includes speech, its accompanying text has no separate charge. Beijing has separate rates; comparable numeric fields cover Singapore text only.

Open profile and sources →
inclusionAI · Text and reasoning

Ling 3.1 Flash 1019

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1M tokens
maximum output
Not published

Weights published under MIT. No API rate has been verified for this revision; numeric costs remain empty. Deployment requires budgeting for memory, accelerators and operations.

Open profile and sources →
NVIDIA · Robotics

Metatron v1

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Motion history and goals
maximum output
29 actions

Model under the NVIDIA Open Model License; ProtoMotions code under Apache 2.0, with separate data and asset rights. There is no verified per-token rate; simulation, hardware and operations have their own costs.

Open profile and sources →
Tencent · Voice and audio

Hy-SimulMT-30B-A3B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Not published

Weights under Apache 2.0, with FP8 and GGUF variants linked as formats of the same family. No hosted API price or universal context or duration limit has been verified. Operation requires resources for the model, audio transport and optional TTS. Voice cloning requires consent and rights to the voice used.

Open profile and sources →
Tencent · Voice and audio

Hy-SimulMT-30B-A3B-Interleave

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Not published

Weights under Apache 2.0, with FP8 and GGUF variants linked as formats of the same family. No hosted API price or universal context or duration limit has been verified. Operation requires resources for the model, audio transport and optional TTS. Voice cloning requires consent and rights to the voice used.

Open profile and sources →
OpenAI · Text and reasoning

GPT‑6.1 Terra

Standard input
1.00 USD / 1 M tokens
Standard output
5.00 USD / 1 M tokens
Cached input
USD 0.05 / 1M tokens
Batch input
0.50 USD / 1 M tokens
Batch output
2.50 USD / 1 M tokens
Context
1,050,000 tokens
maximum output
128,000 tokens

Standard prices per million tokens: USD 1 input, USD 0.05 cached input and USD 5 output. Cache writes cost USD 1.25 for 30 minutes and USD 2.50 for 12 hours. Above 272,000 input tokens, the entire request is charged at 2× input and cache rates and 1.5× output rates. Batch and Flex cost 50% less; Fast costs twice as much. Regional processing adds 10%, and tools may incur additional charges.

Open profile and sources →
inclusionAI · Vision and documents

LBS-27B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Not published

The weights and project declare Apache 2.0; third-party datasets and screenshots retain their own terms. No official API price has been verified. The example uses four devices, without establishing a universal minimum memory requirement. Costs include inference, cropping and, where applicable, the judge model.

Open profile and sources →
Baidu · Vision and documents

Unlimited OCR 2

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Not published

The card declares MIT and the official repository publishes that license. We have not verified a specific commercial price for Unlimited OCR 2; cloud availability of the previous OCR model does not establish an endpoint for this version. The published test environment uses an 80 GB NVIDIA A800, without demonstrating a minimum requirement.

Open profile and sources →
inclusionAI · Text classification

SingProbe Qwen3.8-Flash-Next

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Depends on the base model and runtime
maximum output
Ten scores per token

No API price is published for the probe. Costs include the base model, memory, runtime and monitoring. The card refers to the base model's license; the Qwen variant retains Qwen Community License 1.0, with specific conditions for commercial services. The GLM variant links to MIT. Public downloads do not remove these obligations.

Open profile and sources →
inclusionAI · Text classification

SingProbe GLM-5.3-Flash

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Depends on the base model and runtime
maximum output
Ten scores per token

No API price is published for the probe. Costs include the base model, memory, runtime and monitoring. The card refers to the base model's license; the Qwen variant retains Qwen Community License 1.0, with specific conditions for commercial services. The GLM variant links to MIT. Public downloads do not remove these obligations.

Open profile and sources →
NVIDIA · Text and reasoning

Nemotron 3.5 Bolt E4B

Standard input
Free prototyping endpoint; commercial rate not published
Standard output
Free prototyping endpoint; commercial rate not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
65,536 tokens shared between input and output
maximum output
The card states up to 64K; this is not 64K in addition to the context

NVIDIA Build offers a free prototyping endpoint with the identifier nvidia/nemotron-3.5-bolt-e4b. A commercial rate, universal quota or permanent free access has not been verified; numeric costs remain empty. Downloadable weights require your own infrastructure. The model uses OpenMDW 1.1 and the Gemma 4 base declares Apache 2.0. DFlash drafters are optional speculative-decoding components, not independent chat models or a guarantee of acceleration on every device.

Open profile and sources →
Recursive · Code

Recursive Flint

Standard input
USD 0 / 1M tokens
Standard output
USD 0 / 1M tokens
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
262,144 tokens shared between input and output
maximum output
Configurable output; OpenRouter lists up to 235,929 tokens, including reasoning

OpenRouter lists recursive/flint-v1.0:free at USD 0 for input and output, subject to usage limits; we do not guarantee that it will remain free or that EU in-region routing is available. Self-hosting does incur costs. The files occupy approximately 80 GB for BF16, 22.6 GB for INT4 MLX, 26.0 GB for INT4 Q4_1 GGUF and 22.9 GB for NVFP4 GGUF. These are not total memory requirements: add the KV cache and runtime; the GGUF guides recommend at least 48 GB.

Open profile and sources →
Mistral · Voice and audio

Voxtral Small Transcribe 3

Standard input
0.004 USD / min
Standard output
Included in the audio rate
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Check limits by modality
maximum output
Transcribed text

Public rates in dollars per minute of audio or per million characters, depending on the variant; they are not token prices. Numerical fields per million tokens are left blank to avoid false comparisons. Confirm billing conditions and availability before integrating. Private enterprise deployment requires a specific agreement; no downloadable weights have been published for these variants.

Open profile and sources →
Mistral · Voice and audio

Voxtral Mini Transcribe 3

Standard input
0.0035 USD / min
Standard output
Included in the audio rate
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Check limits by modality
maximum output
Transcribed text

Public rates in dollars per minute of audio or per million characters, depending on the variant; they are not token prices. Numerical fields per million tokens are left blank to avoid false comparisons. Confirm billing conditions and availability before integrating. Private enterprise deployment requires a specific agreement; no downloadable weights have been published for these variants.

Open profile and sources →
Mistral · Voice and audio

Voxtral Mini Transcribe Realtime 3

Standard input
0.008 USD / min
Standard output
Included in the audio rate
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Audio session via WebSocket
maximum output
Transcribed text

Public rates in dollars per minute of audio or per million characters, depending on the variant; they are not token prices. Numerical fields per million tokens are left blank to avoid false comparisons. Confirm billing conditions and availability before integrating. Private enterprise deployment requires a specific agreement; no downloadable weights have been published for these variants.

Open profile and sources →
Mistral · Voice and audio

Voxtral TTS 3

Standard input
13 USD / million characters
Standard output
Included in the text rate
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Text and optional voice reference
maximum output
Synthesized audio

Public rates in dollars per minute of audio or per million characters, depending on the variant; they are not token prices. Numerical fields per million tokens are left blank to avoid false comparisons. Confirm billing conditions and availability before integrating. Private enterprise deployment requires a specific agreement; no downloadable weights have been published for these variants.

Open profile and sources →
NVIDIA · Robotics

Cosmos3-Nano-Sim-Transfer

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Depth, segmentation, edge, or blur controls
maximum output
Up to 600 frames at 480p; playback at 30 fps

Public 16B parameter checkpoint for self-hosted infrastructure. We have not verified a commercial endpoint or token rates: numerical costs are left blank. Downloadable weights do not make inference free. Budget for GPU, memory, video encoding, and operation; review the license of each variant and its dependencies. Playback frames per second do not equate to generation performance.

Open profile and sources →
NVIDIA · Robotics

Cosmos3-Nano-Sim-Bimanual

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Initial image, actions, and camera trajectory
maximum output
Video at 480p; playback at 30 fps

Public 16B parameter checkpoint for self-hosted infrastructure. We have not verified a commercial endpoint or token rates: numerical costs are left blank. Downloadable weights do not make inference free. Budget for GPU, memory, video encoding, and operation; review the license of each variant and its dependencies. Playback frames per second do not equate to generation performance.

Open profile and sources →
NVIDIA · Robotics

Cosmos3-Nano-Transfer-Auto

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Text, WSM controls, and optional visual reference
maximum output
Up to 11 cameras; 300 frames per view; 480p or 720p

Public 16B parameter checkpoint for self-hosted infrastructure. We have not verified a commercial endpoint or token rates: numerical costs are left blank. Downloadable weights do not make inference free. Budget for GPU, memory, video encoding, and operation; review the license of each variant and its dependencies. Playback frames per second do not equate to generation performance.

Open profile and sources →
NVIDIA · Robotics

Cosmos3-Nano-Sim-Auto

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Text, WSM controls, and optional visual reference
maximum output
Up to 11 cameras; 480p or 720p; playback from 10 to 30 fps

Public 16B parameter checkpoint for self-hosted infrastructure. We have not verified a commercial endpoint or token rates: numerical costs are left blank. Downloadable weights do not make inference free. Budget for GPU, memory, video encoding, and operation; review the license of each variant and its dependencies. Playback frames per second do not equate to generation performance.

Open profile and sources →
NVIDIA · Robotics

Cosmos3-H-Embodied-Nano

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
2.4-second clips; up to 16 previous actions
maximum output
25 frames and 24 action steps at 10 Hz, depending on the mode

Public 16B parameter checkpoint for self-hosted infrastructure. We have not verified a commercial endpoint or token rates: numerical costs are left blank. Downloadable weights do not make inference free. Budget for GPU, memory, video encoding, and operation; review the license of each variant and its dependencies. Playback frames per second do not equate to generation performance.

Open profile and sources →
NVIDIA · Image

CDF Sim2Real SMT Solderlight

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
One RGB PCB component image and a fixed English instruction
maximum output
RGB PNG image; target area of 262,144 pixels

Public weights under OpenMDW 1.1; original Qwen-Image-Edit components retain Apache 2.0. The model card estimates about 20B parameters and 170M trained, fused LoRA parameters. We have verified neither a commercial API nor image or token pricing: numeric costs remain empty. Budget for GPU, memory, storage and self-hosted operations.

Open profile and sources →
NVIDIA · Embeddings

Nemotron 3.5 Embed 1B BF16

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 tokens shared between text and image
maximum output
L2-normalized 2,048-dimensional vector

Public, ungated weights under OpenMDW 1.1, with commercial use subject to its terms and preservation of notices. The Ministral bases retain their Apache 2.0 terms. No hosted commercial API, identifier or price has been verified: numeric cost fields remain empty. Downloading does not eliminate GPU, memory and operating costs.

Open profile and sources →
NVIDIA · Embeddings

Nemotron 3.5 Embed 8B BF16

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 tokens shared between text and image
maximum output
L2-normalized 4,096-dimensional vector

Public, ungated weights under OpenMDW 1.1, with commercial use subject to its terms and preservation of notices. The Ministral bases retain their Apache 2.0 terms. No hosted commercial API, identifier or price has been verified: numeric cost fields remain empty. Downloading does not eliminate GPU, memory and operating costs.

Open profile and sources →
NVIDIA · Reranking

Nemotron 3.5 Rerank 8B BF16

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 tokens shared between text and image
maximum output
One relevance logit per query-document pair

Public, ungated weights under OpenMDW 1.1, with commercial use subject to its terms and preservation of notices. The Ministral bases retain their Apache 2.0 terms. No hosted commercial API, identifier or price has been verified: numeric cost fields remain empty. Downloading does not eliminate GPU, memory and operating costs.

Open profile and sources →
NVIDIA · OCR and documents

Nemotron 3.5 Parse

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
16,384 positions shared between image and text
maximum output
9,000 configured tokens; evaluation with 8,744

Downloadable weights under OpenMDW 1.1, with commercial and non-commercial use subject to its terms. The tokenizer has MIT attribution and dependencies retain their licenses. We have not verified a commercial API or token pricing: numeric costs remain empty. Self-hosted inference requires a budget for hardware, memory, operations and security; availability for download does not imply free inference.

Open profile and sources →
NVIDIA · Text and reasoning

Nemotron Labs Dory 1.0 4B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
262,144 tokens
maximum output
Not published

Weights are available in BF16, MXFP8, NVFP4 and GGUF Q5_K_M under OpenMDW 1.1, with commercial and non-commercial use subject to its terms. We have not verified a hosted API tariff: numeric cost fields remain empty. Downloads do not remove hardware, memory, operating and security costs.

Open profile and sources →
Mistral AI · Time series

Timestral 5M

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 time steps
maximum output
1,024 steps per pass · 21 quantiles

No public API tariff has been verified for these checkpoints. Downloads do not cover infrastructure or maintenance. License restrictions also apply to derivatives: review the official text before commercial deployment.

Open profile and sources →
Mistral AI · Time series

Timestral 28M

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 time steps
maximum output
1,024 steps per pass · 21 quantiles

No public API tariff has been verified for these checkpoints. Downloads do not cover infrastructure or maintenance. License restrictions also apply to derivatives: review the official text before commercial deployment.

Open profile and sources →
Mistral AI · Time series

Timestral 123M

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 time steps
maximum output
1,024 steps per pass · 21 quantiles

No public API tariff has been verified for these checkpoints. Downloads do not cover infrastructure or maintenance. License restrictions also apply to derivatives: review the official text before commercial deployment.

Open profile and sources →
Mistral AI · Time series

Timestral 209M

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 time steps
maximum output
1,024 steps per pass · 21 quantiles

No public API tariff has been verified for these checkpoints. Downloads do not cover infrastructure or maintenance. License restrictions also apply to derivatives: review the official text before commercial deployment.

Open profile and sources →
Black Forest Labs · Image

FLUX 3 Image Klein

Standard input
No per-token price
Standard output
USD 0.008–0.036 / image
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Up to ten reference images
maximum output
Up to 4K

The Distilled API charges per image: USD 0.008 at 768sq; 0.009 at 1K; 0.011 at 1.5K; 0.014 at 2K and 0.036 at 4K. No commercial Base endpoint has been verified. These are not token tariffs or local execution costs.

Open profile and sources →
Black Forest Labs · Image

FLUX 3 Image Klein Base

Standard input
No per-token price
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Up to ten reference images
maximum output
Up to 4K

The Distilled API charges per image: USD 0.008 at 768sq; 0.009 at 1K; 0.011 at 1.5K; 0.014 at 2K and 0.036 at 4K. No commercial Base endpoint has been verified. These are not token tariffs or local execution costs.

Open profile and sources →
Meta · Multimodal

Muse Spark Open

Standard input
Depends on the provider
Standard output
Depends on the provider
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
Combined context: 1,000,000 tokens

No single tariff has been verified: hosted APIs belong to third-party providers. The guide states about 700 GB of GPU memory for NVFP4 and 2 TB for BF16. Downloads do not remove infrastructure, operations or safety-review costs.

Open profile and sources →
Liquid AI · Decisions

d1-ultra

Standard input
USD 0.042 / 1M tokens
Standard output
No output tokens
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Probabilities; no generated tokens

Official rate: USD 0.042 per million input tokens, with no output tokens. Each question bills the text and all images again. A 1,024 × 1,024 image uses 1,536 tokens under the published formula. Vercel and OpenRouter are announced as upcoming, not verified as available.

Open profile and sources →
Ai2 · Time series

HiRO-ACEv2

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Climate fields at 100 km; not a token window
maximum output
3 km fields at six-hour intervals

Weights under Apache 2.0, with no verified commercial endpoint or token pricing. Budget for infrastructure, scientific field storage and data preparation. Reported cost savings against X-SHiELD compare inference under Ai2’s protocol; they are not a dollar price or a guarantee for every deployment. The faster simplified version and new variables described as future work are not presented as available.

Open profile and sources →
xAI · Voice and audio

Grok Voice 3.0

Standard input
USD 0.08 / conversation minute
Standard output
USD 0.004 / text input
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Context window not published
maximum output
Streaming audio and text; maximum not published

The official rate is USD 0.08 per conversation minute, equivalent to USD 4.80 per hour, and USD 0.004 per text input. Voice 2.0 pricing is unchanged. These are not rates per million tokens: numerical token-comparison fields remain empty. Separate transcription and speech-synthesis services use different units and are not imported into this profile.

Open profile and sources →
Microsoft · Text and reasoning

UserRepGen-4B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
8,192 prompt tokens; 9,216-token training sequence
maximum output
Up to 1,024 tokens in documented use

The model card declares an MIT licence for the model and retains Apache 2.0 for original Qwen assets. No commercial API or token rate has been verified. Open weights still require compute, memory and maintenance; the CPU smoke test is not a performance measurement. Pin a revision and review licences and permission to process conversations before deployment.

Open profile and sources →
Microsoft · Code

AesCode 8B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
24,576 tokens in the published evaluation
maximum output
12,000 tokens in the published evaluation

The weights declare Apache 2.0; the official repository's code uses MIT. These are different licenses for different artifacts. We have not verified a hosted commercial API or token prices: numeric fields remain empty. Budget for GPU, memory, rendering and operations. The reference-creation pipeline may use external image and vision providers whose costs and terms are not included in the checkpoint.

Open profile and sources →
Microsoft · Code

AesCode 32B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
24,576 tokens in the published evaluation
maximum output
12,000 tokens in the published evaluation

The weights declare Apache 2.0; the official repository's code uses MIT. These are different licenses for different artifacts. We have not verified a hosted commercial API or token prices: numeric fields remain empty. Budget for GPU, memory, rendering and operations. The reference-creation pipeline may use external image and vision providers whose costs and terms are not included in the checkpoint.

Open profile and sources →
Alibaba Qwen · Image

Qwen Image 2.1 Turbo

Standard input
No per-token price
Standard output
USD0.016 / image (Singapore)
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Recommended prompt: up to 4,500 tokens
maximum output
API: 1 to 6 images

Official output-image price: USD0.016 in Singapore and USD0.014133 in Beijing, Hong Kong, Frankfurt, Virginia and Tokyo. These are not per-million-token prices; numeric token-cost fields remain empty. Downloads do not include GPU costs either. Quotas and concurrency depend on region and synchronous or asynchronous mode: confirm your account's operational limit before planning production.

Open profile and sources →
Mistral · Text classification

LIDstral Arabic

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Nonempty text after preprocessing
maximum output
Label and probabilities over 51 classes

The model card declares Apache 2.0, and the repository is public and ungated at this review. Its example still mentions a private download; we do not interpret that wording as an available commercial API. No hosted endpoint or per-token price has been verified. Self-hosting entails CPU, memory, storage and operational costs; pin the code revision before loading model files.

Open profile and sources →
Tencent · Multimodal

Youtu-Parsing-Omni

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
65,536 tokens in the serving example
maximum output
Not published

No hosted commercial API or public per-token price has been verified for this checkpoint. Numeric cost fields remain empty. Self-hosting requires GPUs, dependencies and the provider's vLLM plugin; publicly available files remove neither costs nor territorial restrictions. Evaluation code is listed as awaiting publication, so full harness reproduction is not promised.

Open profile and sources →
Ai2 · Text and reasoning

Bwen 8B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 internal positions; variable byte length
maximum output
max_new_tokens counts bytes; maximum not published

We have not verified a commercial endpoint or an official per-token rate. Numeric cost fields remain empty: downloadable weights do not mean free inference. Budget for memory, compute and maintenance. The documentation uses Python 3.11, Transformers 4.57.3 and xlstm 2.0.4. Bwen and Bolmo Stage 1 declare Apache 2.0; Blama retains Llama 3 Community License, with its own conditions. Also review Ai2’s responsible use guidelines.

Open profile and sources →
Ai2 · Text and reasoning

Blama 8B

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
8,192 internal positions; variable byte length
maximum output
max_new_tokens counts bytes; maximum not published

We have not verified a commercial endpoint or an official per-token rate. Numeric cost fields remain empty: downloadable weights do not mean free inference. Budget for memory, compute and maintenance. The documentation uses Python 3.11, Transformers 4.57.3 and xlstm 2.0.4. Bwen and Bolmo Stage 1 declare Apache 2.0; Blama retains Llama 3 Community License, with its own conditions. Also review Ai2’s responsible use guidelines.

Open profile and sources →
Ai2 · Text and reasoning

Bolmo 1B Stage 1

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
4,096 internal positions; variable byte length
maximum output
max_new_tokens counts bytes; maximum not published

We have not verified a commercial endpoint or an official per-token rate. Numeric cost fields remain empty: downloadable weights do not mean free inference. Budget for memory, compute and maintenance. The documentation uses Python 3.11, Transformers 4.57.3 and xlstm 2.0.4. Bwen and Bolmo Stage 1 declare Apache 2.0; Blama retains Llama 3 Community License, with its own conditions. Also review Ai2’s responsible use guidelines.

Open profile and sources →
Ai2 · Text and reasoning

Bolmo 7B Stage 1

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
65,536 internal positions; variable byte length
maximum output
max_new_tokens counts bytes; maximum not published

We have not verified a commercial endpoint or an official per-token rate. Numeric cost fields remain empty: downloadable weights do not mean free inference. Budget for memory, compute and maintenance. The documentation uses Python 3.11, Transformers 4.57.3 and xlstm 2.0.4. Bwen and Bolmo Stage 1 declare Apache 2.0; Blama retains Llama 3 Community License, with its own conditions. Also review Ai2’s responsible use guidelines.

Open profile and sources →
Ai2 · Text and reasoning

Bwen 8B Stage 1

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
32,768 internal positions; variable byte length
maximum output
max_new_tokens counts bytes; maximum not published

We have not verified a commercial endpoint or an official per-token rate. Numeric cost fields remain empty: downloadable weights do not mean free inference. Budget for memory, compute and maintenance. The documentation uses Python 3.11, Transformers 4.57.3 and xlstm 2.0.4. Bwen and Bolmo Stage 1 declare Apache 2.0; Blama retains Llama 3 Community License, with its own conditions. Also review Ai2’s responsible use guidelines.

Open profile and sources →
Ai2 · Text and reasoning

Blama 8B Stage 1

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
8,192 internal positions; variable byte length
maximum output
max_new_tokens counts bytes; maximum not published

We have not verified a commercial endpoint or an official per-token rate. Numeric cost fields remain empty: downloadable weights do not mean free inference. Budget for memory, compute and maintenance. The documentation uses Python 3.11, Transformers 4.57.3 and xlstm 2.0.4. Bwen and Bolmo Stage 1 declare Apache 2.0; Blama retains Llama 3 Community License, with its own conditions. Also review Ai2’s responsible use guidelines.

Open profile and sources →
Google · Text and reasoning

Gemini 4 Argon

Standard input
2,00 USD / 1 M tokens
Standard output
10.00 USD / 1 M tokens
Cached input
0,10 USD / 1 M tokens
Batch input
1.00 USD / 1 M tokens
Batch output
5.00 USD / 1 M tokens
Context
1,048,576 tokens shared by input, reasoning and output
maximum output
65,536 tokens by default; up to the remaining window with long decode

Introductory USD prices per million tokens through November 30, 2026: input 2, output 10 including thinking, and implicit caching 0.10. From December 1 they become 4, 20 and 0.20. Batch and Flex cost half the standard rates; Priority has different prices. Searches incur separate charges after the shared allowance. There is no free tier or access through Google AI Pro. Numeric fields reflect standard and Batch rates in effect when verified.

Open profile and sources →
Anthropic · Text and reasoning

Claude Haiku 5.5

Standard input
0.10 / 0.50 USD / 1 M tokens
Standard output
0.50 / 2.50 USD / 1 M tokens
Cached input
0.01 / 0.05 USD / 1 M tokens
Batch input
0.05 / 0.25 USD / 1 M tokens
Batch output
0.25 / 1.25 USD / 1 M tokens
Context
1,000,000 tokens
maximum output
128,000 tokens

Each pair of rates corresponds to prompts up to 100,000 tokens / over 100,000 tokens, including output and caching. Input: USD 0.10 / 0.50; output: USD 0.50 / 2.50 per million tokens. Batch discounts input and output by 50%. Caching distinguishes reads from five-minute or one-hour writes. Numeric fields represent only the short-prompt tier, not a long-context budget. Tools and regional conditions may add charges.

Open profile and sources →
Google · Image

Gemini Nano Banana 2.1

Standard input
1,50 USD / 1 M tokens
Standard output
0.0336 / 0.0504 / 0.113 USD per 1K / 2K / 4K image
Cached input
Not published
Batch input
0.75 USD / 1M tokens
Batch output
0.0168 / 0.0252 / 0.0567 USD per 1K / 2K / 4K image
Context
131,072 input tokens
maximum output
32,768 tokens; 1K, 2K and 4K images

Standard rates in USD per million tokens: input 1.50, output text and thinking 7.50, output images 30. Per-image amounts depend on resolution and exclude input, text and searches. Batch halves these rates; Google lists 0.0567 USD per 4K image in Batch. Searches share 5,000 free monthly requests with Gemini 3.x, then cost 14 USD per thousand queries. There is no free API tier. Numeric output fields are left empty to avoid mixing text and images.

Open profile and sources →
IBM · Time series

Granite Timeseries Ensemble R1

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Depends on each member model
maximum output
Nine quantiles from 0.1 to 0.9; median as the point forecast

The identifier refers to a downloadable recipe, not a verified commercial API endpoint. There is no official per-token rate; keep cost unknown and budget for infrastructure. The recipe and implementation are Apache 2.0, but each checkpoint retains its own license. The expanded recipes include IBM Research members under separate, potentially non-commercial terms.

Open profile and sources →
Aleph Alpha · Text and reasoning

Kolibri 1

Standard input
No published official API rate
Standard output
Self-hosted infrastructure cost
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,048,576 tokens; recommended ≤262,144
maximum output
Not published

No official public per-token rate or commercially hosted endpoint has been verified. Downloading permissively licensed weights does not make inference free: budget for GPUs, memory, storage, operations and power. Aleph-Alpha/Kolibri-1 is the repository and self-hosted server identifier in the official example, not a promise of availability on a third-party API.

Open profile and sources →
Alibaba Qwen · Image

Qwen Image 2.1 Pro

Standard input
Not published
Standard output
0.04 USD / image (Singapore)
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Not published

Official price per output image: 0.04 USD in Singapore and 0.035333 USD in Beijing, Hong Kong, Frankfurt, Virginia and Tokyo. Singapore's free quota is ten images and is subject to conditions. This is not a per-token rate; check regional prices and availability before budgeting.

Open profile and sources →
Cohere · Embeddings

Cohere Embed 5 Pro

Standard input
0.12 USD / 1M tokens
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128,000 tokens
maximum output
256, 512, 768, 1,024, 1,536 or 2,048 dimensions

Text costs 0.12 USD per million tokens for Pro and 0.08 USD for Fast; images cost 0.40 USD for both. Model Vault costs depend on the instance. No Batch discount has been verified.

Open profile and sources →
Cohere · Embeddings

Cohere Embed 5 Fast

Standard input
0.08 USD / 1M tokens
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
128,000 tokens
maximum output
256, 512, 768, 1,024, 1,536 or 2,048 dimensions

Text costs 0.12 USD per million tokens for Pro and 0.08 USD for Fast; images cost 0.40 USD for both. Model Vault costs depend on the instance. No Batch discount has been verified.

Open profile and sources →
OpenAI · Text and reasoning

GPT‑6.1 Sol

Standard input
2,00 USD / 1 M tokens
Standard output
10.00 USD / 1 M tokens
Cached input
0,10 USD / 1 M tokens
Batch input
1.00 USD / 1 M tokens
Batch output
5.00 USD / 1 M tokens
Context
1,050,000 tokens · maximum 922,000 input tokens
maximum output
128,000 tokens

Standard prices apply up to 272,000 input tokens. Above that threshold, the entire request costs twice as much for input and cache and 50% more for output. Fast costs twice as much; Batch and Flex reduce prices by 50%. Regional processing adds 10% where available.

Open profile and sources →
Anthropic · Text and reasoning

Claude Opus 5.5

Standard input
4,00 USD / 1 M tokens
Standard output
20,00 USD / 1 M tokens
Cached input
0,20 USD / 1 M tokens
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Anthropic · Text and reasoning

Claude Sonnet 5.5

Standard input
2,00 USD / 1 M tokens
Standard output
10.00 USD / 1 M tokens
Cached input
0,10 USD / 1 M tokens
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
128,000 tokens

Since October 7, 2026, cache reads cost USD 0.10 per million tokens, previously 0.20. Standard input and output remain USD 2.00 and 10.00; Batch remains 1.00 and 5.00. Cache writes cost USD 2.50 for five minutes or 4.00 for one hour. Total savings depend on the share of cached tokens and tools; this is not a fixed discount on every task.

Open profile and sources →
xAI · Text and reasoning

Grok 4.7

Standard input
2,00 USD / 1 M tokens
Standard output
6,00 USD / 1 M tokens
Cached input
0.50 USD / 1 M tokens
Batch input
Not published
Batch output
Not published
Context
500.000 tokens
maximum output
No published text limit

Requests exceeding 200.000 tokens are charged at a higher pricing tier.

Open profile and sources →
xAI · Voice and audio

Grok Voice Transcribe 2.0

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Audio via file or streaming
maximum output
Transcription

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Google · Voice and audio

Gemini 3.8 Live

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
131,072 tokens
maximum output
65.536 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Google · Voice and audio

Gemini 3.8 Live Extended Thinking

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
131,072 tokens
maximum output
65.536 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Alibaba Qwen · Text and reasoning

Qwen3.8-Max

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
Not published

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Alibaba Qwen · Text and reasoning

Qwen3.8-Flash

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Alibaba Qwen · Multimodal

Qwen3.8-Omni-Flash

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Not published
maximum output
Not published

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Alibaba Qwen · Voice and audio

Qwen3.8-Omni-Flash-Realtime

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
196.608 input tokens
maximum output
Not published

Consult the primary source before budgeting for a deployment.

Open profile and sources →
ElevenLabs · Voice and audio

Eleven v4

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Voiceover text
maximum output
Audio

Consult the primary source before budgeting for a deployment.

Open profile and sources →
ElevenLabs · Voice and audio

Eleven v4 Turbo

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Voiceover text
maximum output
Streaming audio

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Z.AI · Code

GLM‑5.3

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
128,000 tokens

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Z.AI · Multimodal

GLM‑5.3‑Flash

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
1,000,000 tokens
maximum output
Not published

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Alibaba Qwen · Image

Qwen Image 3.0 Pro

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
4,500 input tokens
maximum output
Up to 6 images · 2,048 × 2,048 pixels

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Alibaba Qwen · Video

Wan 3.0 Video Prime

Standard input
Not published
Standard output
Not published
Cached input
Not published
Batch input
Not published
Batch output
Not published
Context
Multimodal references
maximum output
Videos up to 30 seconds · 1080p

Consult the primary source before budgeting for a deployment.

Open profile and sources →
Cost per task

Includes retries, tooling, recovery, cache, and human review.

Operating limit

Account and region limits may differ from the theoretical capacity of the model.

Missing data

The absence is retained as relevant information, not as a zero.

What is missing from the per-token price

A low rate can end up being expensive if the workflow repeats work or requires extensive oversight.

01

Reasoning and length

Internal or thinking tokens are usually billed as output and depend on difficulty and the configured effort.

02

Tools and grounding

Search, maps, code execution, and other tools may have per-call pricing in addition to tokens.

03

Caching and batches

Repeated prefixes and asynchronous processing reduce cost, but require designing the workflow to take advantage of them.

04

Cost of error

Measure retries, human validation, incidents, and recovery: they are part of the total operating cost.

ANALYSIS

Related analysis

AesCode 8B: Calculating the Cost of Aesthetic Code Generation Tasks
GUIA

AesCode 8B: Calculating the Cost of Aesthetic Code Generation Tasks

The cost of using AesCode 8B locally can include inference, rendering, verification, and retries. This calculation model separates those components and shows how to replace assumptions with your own measurements.

07 Nov 2026
AesCode 32B: Verification Protocol for Local and Self-Hosted Deployment Costs
GUIA

AesCode 32B: Verification Protocol for Local and Self-Hosted Deployment Costs

An operational guide to auditing the real costs of AesCode 32B, distinguishing between technical specifications confirmed by Microsoft and economic variables that require internal estimation, given that no public API pricing exists for this model.

18 Oct 2026
Claude Opus 5.5 in production: how to estimate cost per task
COMPARATIVA

Claude Opus 5.5 in production: how to estimate cost per task

The model-specific documentation lists 4 and 20 US dollars for Claude Opus 5.5 but does not confirm the billing unit. This guide provides a formula, conditional scenarios, and a method for measuring your own workload, without presenting cache or batch rates as confirmed.

17 Oct 2026
Cohere Embed 4: How to Measure the Cost of a Useful Retrieval
COMPARATIVA

Cohere Embed 4: How to Measure the Cost of a Useful Retrieval

Comparing a per-call or per-use price is not enough. To determine whether Embed 4 is worthwhile for a particular corpus, first verify the commercial terms for the access channel, then measure quality using queries and relevance criteria defined in advance.

01 Oct 2026
Claude Sonnet 4.5 with tools: calculate the cost of a task, not just a call
GUIA

Claude Sonnet 4.5 with tools: calculate the cost of a task, not just a call

A template for estimating tool-workflow costs by round, separating input and output tokens from additional charges, and checking the calculation against usage logs.

28 Sep 2026
Claude Opus 4.5 for Long Tasks: Calculate the Cost per Completed Job
GUIA

Claude Opus 4.5 for Long Tasks: Calculate the Cost per Completed Job

A price per million tokens is not enough to budget a multi-step workflow. Learn how to add up input, output, cache, retries, and review, and set limits before you run it.

25 Sep 2026