# Kimi K2.5 (image)

Frontier-scale agentic work with images. Moonshot AI's 1T-parameter multimodal agentic MoE with image input and function calling. Served on GraphN managed inference behind an OpenAI-compatible API as `kimi-k2.5`.

## Overview

Kimi K2.5 is Moonshot AI's open-source native-multimodal agentic model: a 1T-parameter mixture of experts with 32B active per token, continually pretrained on roughly 15T mixed vision and text tokens on top of Kimi-K2-Base, with a 400M-parameter MoonViT vision encoder. Moonshot highlights generating code from visual specifications and splitting a task across dynamically created sub-agents.

## Strengths

- By far the largest model in the catalogue
- Native image understanding
- Trained for multimodal agentic work

## Use cases

- Frontier-scale agentic tasks with many tool calls
- Code generation from screenshots and visual specs
- Long-horizon plans that mix text and images

## Good to know

Moonshot's thinking mode is not exposed on this alias, and video input is not supported.

## Specifications

- **Type**: Chat
- **Context**: 256K context
- **Parameters**: 1T total, 32B active (MoE)
- **License**: Modified MIT
- **Family**: Kimi K2
- **Weights**: moonshotai/Kimi-K2.5
- **Pricing**: Input $0.60/1M · Output $3.00/1M

## API access

### curl

```bash
curl https://model.graphn.ai/v1/chat/completions \
  -H "Authorization: Bearer $GRAPHN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k2.5",
    "messages": [{"role": "user", "content": "Hello"}]
  }'
```

### CLI

```bash
graphn model chat kimi-k2.5 -m "What is GraphN?"
```

### Python

```python
@function
async def ask(question: str) -> str:
    return await chat.complete(question, model="kimi-k2.5")
```

[Start free](https://graphn.ai/signup?next=%2Finference%2Fkimi-k2.5)

[Model card](https://huggingface.co/moonshotai/Kimi-K2.5) · [Managed inference](https://graphn.ai/inference) · [Pricing](https://graphn.ai/pricing) · [API reference](https://graphn.ai/api)