123 models with profiles. Classification describes capabilities, not a quality ranking.
NVIDIA image-editing model, version 1.0.0 released on October 20, 2026. It transforms synthetic PCB component crops rendered in Omniverse into an optical-inspection lighting style. It derives from Qwen-Image-Edit with a fused LoRA fine-tune: the package includes the complete pipeline, not a separate adapter. Users supply the image; the English instruction is fixed, not a free-form prompt.
Image · Open weights↗ NVIDIANVIDIA embedding model with 1.55B parameters, released on October 20, 2026. It vectorizes text documents, page images or both, with text-only queries. Distilled from Embed 8B, it freezes the Nemotron 3 Embed 1B text tower and adds visual retrieval. Its vector has 2,048 dimensions; it does not automatically share the 8B model’s space.
Embeddings · Open weights↗ NVIDIANVIDIA embedding model with 8.38B parameters, released on October 20, 2026. It vectorizes text, page images or both for retrieval with text queries. It freezes the Nemotron 3 Embed 8B tower and adds vision through Pixtral. It retains that predecessor’s text embedding space; it produces 4,096-dimensional vectors, not conversational responses.
Embeddings · Open weights↗ NVIDIANVIDIA reranker with 8.38B parameters, released on October 20, 2026. It scores each text-query and document pair; documents may contain text, a page image or both. It complements initial retrieval by reranking candidates; it does not generate responses or embeddings. It combines a Pixtral vision encoder with a bidirectional Ministral3 tower.
Reranking · Open weights↗ DeepSeekDeepSeek updated its API, website and app to DeepSeek-V4.1-Flash-1017 on October 17, 2026. The API identifier remains deepseek-flash and pricing is unchanged. We retain the profile and its URL: this is a revision of the existing model, not another duplicate model.
Text and reasoning · Paid API · open weights↗ NVIDIANVIDIA model released on October 17, 2026 to turn page images into structured text, semantic classes, boxes and reading order. It combines C-RADIOv4-SO400M with a ten-layer mBART decoder and 630,800,496 unique parameters. It accepts an RGB image and a task prompt; PDFs must be converted into page images. It does not generate images or replace a general-purpose assistant.
OCR and documents · Open weights↗ NVIDIANVIDIA text model released on October 16, 2026, with 4 billion parameters and a recurrent architecture. It reuses a block of layers to adjust compute per token without switching checkpoints. English and a validated 256K-token context are documented. It does not generate images, audio or video.
Text and reasoning · Open weights↗ Liquid AILiquid AI multimodal decision model released on October 15, 2026. It accepts text and images and returns probabilities for predefined choices, binary questions or scales. It uses the d1-ultra identifier in the decisions API; it does not generate free-form text.
Decisions · Paid proprietary↗ Black Forest LabsA 5.6-billion-parameter image model released on October 15, 2026. It generates and edits images from text and references, with bounding-box composition control. Distilled and Base are different checkpoints; neither should be confused with the full FLUX 3 Image model.
Image · Restricted weights↗ Black Forest LabsA 5.6-billion-parameter image model released on October 15, 2026. It generates and edits images from text and references, with bounding-box composition control. Distilled and Base are different checkpoints; neither should be confused with the full FLUX 3 Image model.
Image · Restricted weights↗ MetaMeta multimodal model released on October 15, 2026, with 816 billion total and 42 billion active parameters. It accepts text, images, audio and video and returns text. It has a combined one-million-token context and BF16 and NVFP4 variants.
Multimodal · Restricted weights↗ Mistral AIMistral numerical model family released on October 15, 2026 in four sizes: 5M, 28M, 123M and 209M. It performs zero-shot univariate and multivariate probabilistic forecasting. It accepts history, past covariates and known future covariates; it neither chats nor generates text.
Time series · Restricted weights↗ Mistral AIMistral numerical model family released on October 15, 2026 in four sizes: 5M, 28M, 123M and 209M. It performs zero-shot univariate and multivariate probabilistic forecasting. It accepts history, past covariates and known future covariates; it neither chats nor generates text.
Time series · Restricted weights↗ Mistral AIMistral numerical model family released on October 15, 2026 in four sizes: 5M, 28M, 123M and 209M. It performs zero-shot univariate and multivariate probabilistic forecasting. It accepts history, past covariates and known future covariates; it neither chats nor generates text.
Time series · Restricted weights↗ Mistral AIMistral numerical model family released on October 15, 2026 in four sizes: 5M, 28M, 123M and 209M. It performs zero-shot univariate and multivariate probabilistic forecasting. It accepts history, past covariates and known future covariates; it neither chats nor generates text.
Time series · Restricted weights↗ xAIProprietary voice model available since October 14, 2026 for bidirectional audio and text conversations. Its versioned identifier is grok-voice-3.0; grok-voice-latest is an alias already pointing to this version, not another model. It replaces the deprecated Grok Voice 2.0 without changing the endpoint, session parameters or events. Do not confuse it with Grok Voice Transcribe 2.0 or the separate text-to-speech service.
Voice and audio · Paid proprietary↗ Ai2Ai2 system announced on October 14, 2026 combining the ACE2S-X-SHiELD atmospheric emulator with the HiROv2 diffusion model. It transforms roughly 100 km fields into 3 km precipitation, ten-metre wind and sea-level pressure data. The public repository includes both checkpoints, configuration and sample data; its creation on October 12 is not confused with the announcement date. It is not a conversational assistant or a conventional image generator.
Time series · Open weights↗ MicrosoftMicrosoft checkpoint based on Qwen3.5-4B that turns text conversations into descriptions of the task and observed behaviour. Weights were uploaded to the official repository on October 14, 2026. It generates representations for a separate simulator: it does not itself produce user turns or execute their task. Although it retains the base model’s multimodal architecture, this release supports text only and was evaluated in English.
Text and reasoning · Open weights↗
Technical analysis of NVIDIA's model for converting simulation controls into RGB video. We evaluate its architecture, published benchmarks, hardware requirements, and the physical limitations that prevent its use as a certified simulator.
22 Oct 2026 ↗
A rigorous evaluation of Mistral's text-to-speech model: hybrid architecture, evaluation metrics, per-character pricing, and practical limitations for engineering teams.
22 Oct 2026 ↗
A technical evaluation of Mistral's offline transcription variant: architecture, provider-reported metrics, deployment requirements, and criteria for choosing it over alternatives.
22 Oct 2026 ↗
Z.AI presents GLM-5.1 as a model for agentic engineering, and a model card for it is listed on Hugging Face. But the available information is not enough to confirm its license, execution requirements, official pricing, or benchmark results. This guide separates what is documented from what remains unverified and offers a checklist for evaluating the model.
03 Oct 2026 ↗
Cohere says Embed 5 Pro achieved the highest average score among the models it tested and highlights its performance on financial document retrieval. The available information does not make it possible to verify the score, methodology, exact price, or all conditions of use.
03 Oct 2026 ↗
The available documentation identifies Embed 5 Fast as part of the Embed 5 family, but it is not enough to confirm its input modalities, availability by channel, applicable price, or retrieval results. Here is what can be stated—and what still needs to be verified.
03 Oct 2026 ↗
Google describes Gemini 3.8 Live Extended Thinking as an audio-to-audio model for the Live API, recommended when more background reasoning is needed. The available documentation does not, by itself, show that this reduces errors or improves latency. Those questions require testing with controlled tasks and conditions.
02 Oct 2026 ↗
Google describes Gemini 3.8 Live as a model for real-time dialogue, but the available sources do not confirm a price, quantitative results, or complete access conditions. This guide separates the provider’s claims from what a team should verify in its own integration.
02 Oct 2026 ↗
ElevenLabs presents v4 Turbo as a low-latency variant and publishes median inference latency of around 100 ms. That figure is not the response time of a complete agent: measurement details, access conditions, applicable pricing, and model-specific security evidence are still missing.
02 Oct 2026 ↗