# Kimi K2.5

> Multimodal agentic. 262K context.

Human version: https://kymaapi.com/models/kimi-k2.5
Live JSON: `GET https://kymaapi.com/v1/models` (no auth required)

## Facts

- **Model ID**: `kimi-k2.5` — pass this as `model` in the request body
- **Creator**: Moonshot AI
- **Released**: 2026-01-26
- **Context window**: 262K tokens
- **Max output**: 33K tokens per response — a hard ceiling, not a default
- **Price**: $0.6075 in / $3.038 out per 1M
- **Capabilities**: tools, vision, reasoning

## Call it

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

## Positioning

Kimi K2.5 is a frontier multimodal model optimized for agentic workflows, tool use, and vision tasks. Developers building autonomous agents or complex coding pipelines should use it when they need reliable structured outputs and long-context reasoning.

## About

Kimi K2.5 is a multimodal language model from Moonshot with a 262,144 token context window. It accepts text and image inputs, outputs text, and includes native support for tool calling, structured outputs, and explicit reasoning traces. It operates in the premium tier and delivers medium inference speed.

On Kyma, the model runs through an OpenAI-compatible endpoint using a single platform API key. Every request receives automatic failover if a serving path degrades, and responses return exact token costs in usage.cost alongside an X-Kyma-Model header. While Kyma bills repeated prompt prefixes at 10% of the input rate, this specific model does not support caching, so all input tokens bill at the full rate. New accounts include a $0.50 free credit with no card required.

The model caps generation at 32,768 output tokens per request. It does not process audio or video inputs, and its medium speed tier makes it unsuitable for strict low-latency requirements. Platform telemetry shows a 99.5% success rate across approximately 34.8 million tokens served.

## Use cases

- **Autonomous Agent Workflows** — Leverages tool calling and structured outputs to coordinate multi-step tasks.
- **Code Generation and Review** — Processes long codebases within its 262K context window for refactoring and debugging.
- **Image Analysis Pipelines** — Extracts structured data and reasoning from uploaded images alongside text prompts.
- **Complex Chat Applications** — Maintains conversational state across extended sessions with reliable reasoning capabilities.

## Not ideal for

Do not use this model for high-throughput, low-latency streaming or tasks that require prompt caching to reduce input costs.

## Pick something else when

- You need faster inference for real-time chat → `gemini-3.5-flash-lite`
- You require prompt caching to lower prefix costs → `qwen-3.7-plus`
- You are optimizing for budget over frontier reasoning → `deepseek-v4-flash`

## See also

- All models: https://kymaapi.com/models.md
- Pricing: https://kymaapi.com/pricing.md
- Other models by Moonshot AI: https://kymaapi.com/models?q=Moonshot%20AI