Primary use stated or inferred cautiously from official documentation.
M
MEITUAN / LongCat-2.5
LongCat-2.5
Meituan multimodal model with 1.6 trillion total parameters and about 48 billion activated per token, designed for agents and visual programming. The stable version has been available through the API since November 2, 2026 and is distinct from LongCat-2.5-Preview. It accepts text, images and video, producing text and code responses.
01 / SUMMARY
The essential
Maximum input capacity when published by the source.
Documented maximum generation limit.
Verified availability channels.
Primary functional family and verified specialties.
Declared terms for API access, model weights, or self-hosting.
02 / TECHNICAL SHEET
Limits and integration
| API ID | LongCat-2.5 |
|---|---|
| Model type | Multimodal · Code · Text and reasoning |
| Access model | Paid proprietary |
| License | No verified weights license |
| Deployment | Hosted API |
| Release | 02 NOV 2026 |
| Knowledge cutoff | Not published |
| Entrance | Text · Image · Video |
| Exit | Text · Code |
| Context window | 1.048.576 tokens |
| maximum output | 262,144 tokens |
| Reasoning | Optional reasoning; maximum budget of 262,144 tokens |
| Published tools | Tool calls |
| Structured Outings | Not published |
| Batch processing | Not published |
| Prompt cache | Cached input with a separate rate |
| Fine-tuning | Not published |
| Verified platforms | LongCat · API compatible with OpenAI and Anthropic |
Repository development, screenshot and chart interpretation, and multi-step tool workflows. The documentation supports integrating the model into OpenAI- or Anthropic-compatible clients. Protocol compatibility does not imply support for every option offered by those providers. Evaluate each integration with bounded tasks and permissions.
The context window includes both input and output. Images allow up to 10 MB per file and 100 per request; videos allow up to 100 MB per file, ten per request and a duration of less than one hour. Video is supplied by URL. These limits do not guarantee reliable understanding of every detail. No weight release for this stable version or image or audio generation capabilities have been verified.
03 / CAPABILITIES
What it can do
Orientation
Repository development, screenshot and chart interpretation, and multi-step tool workflows. The documentation supports integrating the model into OpenAI- or Anthropic-compatible clients. Protocol compatibility does not imply support for every option offered by those providers. Evaluate each integration with bounded tasks and permissions.
Context
1,048,576 tokens · 262,144 tokens
Tools and integration
Tool calls
Access
LongCat website and API
Input modalities
TextImageVideo
04 / PRICES
Documented cost
| Concept | Worth | Unit/condition |
|---|---|---|
| Standard input | 0.30 USD / 1M tokens | Standard API rate |
| Cached input | 0.006 USD / 1M tokens | Reading reused prefixes |
| Cache write or storage | Not published | The condition varies by provider |
| Standard output | 1.20 USD / 1M tokens | May include reasoning tokens |
| Batch input | Not published | Asynchronous processing |
| Batch output | Not published | Asynchronous processing |
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.
i Prices change and may depend on level, region or context length. Check the source before making a decision.
05 / EVALUATIONS
How to read the results
“
Meituan reports 88.4 out of 100 on Terminal Bench v2.1, using Claude Code v2.1.280, a maximum sequence length of 512K and the average of three runs. The general setup enables reasoning with temperature 1.0 and top_p 0.95. This is a provider result not reproduced by Inferama; other models’ figures in the table may come from different protocols. It does not represent the success rate on a particular repository.
| Benchmark | Result | Metric | Source |
|---|---|---|---|
| Terminal-Bench 2.1 | 88,4 / 100 | Score | View source ↗ |
Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.
06 / VERSIONS
Chronology
LongCat-2.5
Stable API release confirmed in the official changelog.
LongCat-2.5 · anuncio técnico
Date of the technical article; API availability is confirmed the following day.
07 / SOURCES
Traceability
TECHNICAL SHEETQuick start and supported API modellongcat.ai↗
LAUNCHStable availability on November 2longcat.ai↗
LIMITSParameters and context limitslongcat.ai↗
MODALIDADESImage and video understandinglongcat.ai↗
PRICESOfficial promotional pricinglongcat.ai↗
ASSESSMENTArchitecture and evaluation protocollongcat.ai↗