# Kimi K2.5

> Moonshot AI's trillion-parameter open-weight multimodal agent model

- Maker: Moonshot AI (Kimi)
- Line: Kimi K2
- Canonical page: https://ai-tldr.dev/models/kimi-k2-5/
- Updated: 2026-06-18

## In short

Kimi K2.5 is Moonshot AI's open-weight flagship model, released on January 27, 2026 as the multimodal successor in the company's Kimi K2 line. It is a Mixture-of-Experts model with 1 trillion total parameters but only 32 billion active per token, which keeps inference cost low relative to its size. It has a 256K-token context window and costs $0.60 per million input tokens and $2.50 per million output tokens.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 256K (max output 16K) | [1] |
| Input price | $0.60 / 1M tokens | [2] |
| Output price | $2.50 / 1M tokens | [2] |
| Licence | Modified MIT License · open weights | [1] |
| Released | 27 January 2026 | [3] |
| Parameters | 1T total, 32B active (MoE) | [1] |

## Sources

1. [Kimi K2.5 on Hugging Face](https://huggingface.co/moonshotai/Kimi-K2.5)
2. [Kimi K2.5 complete guide (Codecademy)](https://www.codecademy.com/article/kimi-k-2-5-complete-guide-to-moonshots-ai-model)
3. [Kimi K2.5 tech blog: Visual Agentic Intelligence](https://www.kimi.com/blog/kimi-k2-5)

## Overview

Kimi K2.5 is Moonshot AI's open-weight flagship model, released on January 27, 2026 as the multimodal successor in the company's Kimi K2 line. It is a Mixture-of-Experts model with 1 trillion total parameters but only 32 billion active per token, which keeps inference cost low relative to its size. Unlike the original text-only Kimi K2, K2.5 is natively multimodal: it was built by continued pretraining over roughly 15 trillion mixed visual and text tokens and pairs the language backbone with a 400M-parameter MoonViT vision encoder.

K2.5 accepts text, images, and (experimentally) video, with a 256K-token context window. Moonshot positions it as a 'visual agentic' model aimed at real work, exposing four operating modes: Instant for quick answers, Thinking for step-by-step reasoning, Agent for research and content creation, and an Agent Swarm mode that can coordinate up to 100 sub-agents across many parallel steps for large-scale research, long-form writing, and batch tasks.

The weights are published on Hugging Face under a Modified MIT License that permits commercial use, so K2.5 can be self-hosted. It is also served through Moonshot's own Kimi API platform and through third parties including Amazon Bedrock and OpenRouter. Note that Moonshot has since shipped newer K2.6 and K2.7 models, so K2.5 is now an earlier release in the line rather than the current flagship.

## Benchmarks

- SWE-bench Verified: 76.8% ([source](https://github.com/MoonshotAI/Kimi-K2.5))
- SWE-bench Multilingual: 73% ([source](https://www.kimi.com/blog/kimi-k2-5))
- SWE-bench Pro: 50.7% ([source](https://huggingface.co/moonshotai/Kimi-K2.5))
- Terminal-Bench 2.0: 50.8% ([source](https://www.kimi.com/blog/kimi-k2-5))
- LiveCodeBench v6: 85% ([source](https://www.kimi.com/blog/kimi-k2-5))
- AIME 2025: 96.1% ([source](https://github.com/MoonshotAI/Kimi-K2.5))
- GPQA-Diamond: 87.6% ([source](https://github.com/MoonshotAI/Kimi-K2.5))
- MMLU-Pro: 87.1% ([source](https://github.com/MoonshotAI/Kimi-K2.5))
- Humanity's Last Exam (full): 30.1% ([source](https://www.kimi.com/blog/kimi-k2-5))
- VideoMMMU: 86.6% ([source](https://www.kimi.com/blog/kimi-k2-5))
- VideoMME: 87.4% ([source](https://github.com/MoonshotAI/Kimi-K2.5))
- MMMU-Pro: 78.5% ([source](https://www.kimi.com/blog/kimi-k2-5))
- OCRBench: 92.3% ([source](https://huggingface.co/moonshotai/Kimi-K2.5))

## Pricing

- Input: $0.60 / 1M tokens per 1M tokens
- Output: $2.50 / 1M tokens per 1M tokens

Standard list price on Moonshot's Kimi API. Third-party providers vary; OpenRouter lists about $0.375 input / $2.025 output per 1M tokens. Self-hosting the open weights is also an option.

Source: https://www.codecademy.com/article/kimi-k-2-5-complete-guide-to-moonshots-ai-model

## Strengths

- Strong agentic coding: 76.8% on SWE-bench Verified and 73.0% on SWE-bench Multilingual, competitive with leading proprietary models
- Open weights under a Modified MIT License that allows commercial use and self-hosting
- Cost-efficient MoE design that activates only 32B of 1T parameters per token
- Native multimodal vision (built on a MoonViT encoder) with strong document, OCR, and video-understanding scores
- Agent Swarm mode that parallelizes work across up to 100 coordinated sub-agents
- Large 256K-token context window for long documents and codebases

## Best for

- Agentic software engineering: multi-file bug fixes, refactors, and repo-level tasks
- Visual coding and turning screenshots or design images into working code
- Long-document analysis, OCR, and structured extraction from PDFs and scans
- Large-scale parallel research and long-form writing via Agent Swarm
- Self-hosted deployment for teams needing on-prem or private inference
- Multilingual coding and reasoning across many programming languages

## How to access

- [Moonshot AI (Kimi)](https://platform.moonshot.ai/docs/introduction) — model id `kimi-k2.5`
- [Amazon Bedrock](https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-5.html) — model id `moonshotai.kimi-k2.5`
- [OpenRouter](https://openrouter.ai/moonshotai/kimi-k2.5) — model id `moonshotai/kimi-k2.5`

## Kimi K2 versions

- [Kimi K2.7-Code](https://ai-tldr.dev/models/kimi-k2-7-code/) (12 June 2026)
- [Kimi K2.6](https://ai-tldr.dev/models/kimi-k2-6/) (20 April 2026)
- Kimi K2.5 (27 January 2026) — this page
- [Kimi K2-Instruct-0905](https://ai-tldr.dev/models/kimi-k2-instruct-0905/) (9 September 2025)
- [Kimi K2](https://ai-tldr.dev/models/kimi-k2/) (11 July 2025)

## FAQ

### Is Kimi K2.5 open source?

The model weights are openly published on Hugging Face under a Modified MIT License that permits commercial use, so it is open-weight and can be self-hosted. Moonshot also offers it as a hosted API.

### How big is Kimi K2.5?

It is a Mixture-of-Experts model with 1 trillion total parameters, but only about 32 billion are activated per token (8 routed experts plus 1 shared, out of 384), which keeps inference far cheaper than a dense trillion-parameter model would be.

### What modalities and context length does Kimi K2.5 support?

It natively handles text and images, with experimental video understanding, via a 400M-parameter MoonViT vision encoder. The context window is 256K tokens. It does not support audio input or output.

### How much does Kimi K2.5 cost to use?

Moonshot's standard list price on the Kimi API is about $0.60 per million input tokens and $2.50 per million output tokens. Third-party providers vary, and you can also run the open weights yourself.

## Links

- [Kimi K2.5 GitHub repository](https://github.com/MoonshotAI/Kimi-K2.5)
- [Kimi K2.5 on Hugging Face](https://huggingface.co/moonshotai/Kimi-K2.5)
- [Kimi K2.5 tech blog: Visual Agentic Intelligence](https://www.kimi.com/blog/kimi-k2-5)
- [Kimi K2.5 product page](https://www.kimi.com/ai-models/kimi-k2-5)
- [Moonshot AI API documentation](https://platform.moonshot.ai/docs/introduction)
- [Kimi K2.5 on Amazon Bedrock](https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-5.html)
- [Kimi K2.5 on OpenRouter](https://openrouter.ai/moonshotai/kimi-k2.5)
- [Kimi K2.5 complete guide (Codecademy)](https://www.codecademy.com/article/kimi-k-2-5-complete-guide-to-moonshots-ai-model)
- [China's Moonshot releases Kimi K2.5 (TechCrunch)](https://techcrunch.com/2026/01/27/chinas-moonshot-releases-a-new-open-source-model-kimi-k2-5-and-a-coding-agent/)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/kimi-k2-5/