# LongCat-2.5

> LongCat-2.5 is Meituan's natively multimodal agentic model: a sparse mixture-of-experts with 1.6T total parameters and about 48B active, over a 1M-token context window. It takes text and image input and returns text. It is built for long-horizon agentic work, agentic coding, and visual coding such as creative design and 3D tasks, and has a `thinking` toggle. Served over the OpenAI-compatible and Anthropic-compatible APIs.

**ID**: `meituan/longcat-2.5`  
**Creator**: meituan  
**Category**: text  
**Context**: 1M tokens  
**Pricing**: $0.30 / $1.18 per 1M · billed by your provider (BYOK)  
**Released**: 2026-11-02  
**Web page**: https://anyrouter.dev/model/meituan/longcat-2.5

**Input modalities**: text, image  
**Output modalities**: text  

**Capabilities**: chat, streaming, reasoning, coding, agentic, function-calling, vision

**Aliases**: `meituan/longcat-2.5-preview`, `longcat/longcat-2.5-preview`, `longcat/longcat-2.5`

## Usage

**Endpoint**: `POST https://anyrouter.dev/api/v1/chat/completions`

```bash
curl https://anyrouter.dev/api/v1/chat/completions \
  -H "Authorization: Bearer $ANYROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "meituan/longcat-2.5",
  "messages": [
    {
      "role": "user",
      "content": "Say hi in 3 words."
    }
  ]
}'
```

## Providers

| Provider | Input | Output |
| --- | --- | --- |
| longcat-byok | $0.30/M (list) | $1.20/M (list) |
| vercel-byok | $0.30/M (list) | $1.20/M (list) |
| zenmux-byok | $0.30/M (list) | $1.18/M (list) |

BYOK routes are billed by your own provider key — AnyRouter charges $0. Rates marked (list) are the provider's published list price.

## Model card

- Architecture: mixture of experts [1]
- Total parameters: 1600B [1]
- Active parameters: 48B [1]
- Native context: 1048576 tokens [3]

### Strengths [4,1]

- Native multimodal input: image understanding paired with agentic coding (vendor-stated).
- Built for long-horizon agent work such as complex codebases and full-project builds (vendor-stated).
- Works across agent harnesses such as Claude Code, Codex and Hermes (vendor-stated).

### Known limits [2]

- Text output only; the API returns text, not images.
- Hugging Face and GitHub weight releases show as "coming soon" on the launch page, so open-weight status and license are not confirmed.

### Use cases [1,4]

- Agentic coding and long-running engineering tasks
- Visual coding such as creative design and 3D tasks
- Image understanding in agent workflows

### API notes [3]

- A `thinking` parameter turns reasoning on or off (`{"type":"enabled"}` / `{"type":"disabled"}`); temperature is fixed at 1.0 with thinking on and 0.7 with it off.
- Context is 1,048,576 tokens; `max_tokens` is capped at 262,144 and `thinking_budget` at 262,144.

### Links

- announcement: https://longcat.ai/blog/longcat-2.5/
- docs: https://longcat.chat/platform/docs/
- pricing: https://longcat.chat/platform/docs/pricing/longcat-2.5
- Previous version: https://anyrouter.dev/model/meituan/longcat-2.0

### Sources

1. [Meituan LongCat: Introducing LongCat-2.5](https://x.com/Meituan_LongCat/status/2117264311714537633) (accessed 2026-11-02)
2. [Introducing LongCat-2.5](https://longcat.ai/blog/longcat-2.5/) (accessed 2026-11-02)
3. [LongCat API: Model Parameters](https://longcat.chat/platform/docs/api/model-params) (accessed 2026-11-02)
4. [LongCat API Platform Change Log](https://longcat.chat/platform/docs/change-log) (accessed 2026-11-02)

## Source

- Upstream docs: https://longcat.ai/blog/longcat-2.5/