Introducing Kimi K2 Thinking
Today, we are introducing Kimi K2 Thinking, our best open-source thinking model.
Built as a thinking agent, it reasons step by step while using tools, achieving state-of-the-art performance on Humanity's Last Exam (HLE), BrowseComp, and other benchmarks, with major gains in reasoning, agentic search, coding, writing, and general capabilities.
Kimi K2 Thinking can execute up to 200 – 300 sequential tool calls without human interference, reasoning coherently across hundreds of steps to solve complex problems.
It marks our latest efforts in test-time scaling, by scaling both thinking tokens and tool calling steps.
K2 Thinking is now live on kimi.com under the chat mode, with its full agentic mode available soon. It is also accessible through the Kimi K2 Thinking API.
Evaluations
Kimi K2 Thinking sets new records across benchmarks that assess reasoning, coding, and agent capabilities. K2 Thinking achieves:
- 44.9% on HLE with tools
- 60.2% on BrowseComp
- 71.3% on SWE-Bench Verified
Demonstrating strong generalization as a state-of-the-art thinking agent model.
评测基准
| 类别 | 基准测试 | 描述 |
|---|---|---|
| Agentic Reasoning | Humanity's Last Exam (Text-only) w/ tools | 专家级跨学科问题 |
| Agentic Search | BrowseComp | 智能体搜索与浏览 |
| Agentic Search | Seal-0 | 真实世界最新信息收集 |
| Coding | SWE-Multilingual | 智能体编码 |
| Coding | SWE-bench Verified | 智能体编码 |
| Coding | LiveCodeBench V6 | 竞争性编程 |
Agentic Reasoning
K2 Thinking demonstrates outstanding reasoning and problem-solving abilities. On Humanity's Last Exam (HLE) — a rigorously crafted, closed-ended benchmark spanning thousands of expert-level questions across more than 100 subjects — K2 Thinking achieved a state-of-the-art score of 44.9%, with search, python, and web-browsing tools, establishing new records in multi-domain expert-level reasoning performance.
By reasoning while actively using a diverse set of tools, K2 Thinking is capable of planning, reasoning, executing, and adapting across hundreds of steps to tackle some of the most challenging academic and analytical problems. In one instance, it successfully solved a PhD-level mathematics problem through 23 interleaved reasoning and tool calls, exemplifying its capacity for deep, structured reasoning and long-horizon problem solving.
Agentic Search
K2 Thinking excels at real-world information gathering and agentic search tasks. On BrowseComp, it achieved 60.2%, demonstrating strong generalization capability in web browsing and information synthesis.
Coding
In coding benchmarks, K2 Thinking shows impressive performance:
- 71.3% on SWE-Bench Verified
- Strong results on SWE-Multilingual and LiveCodeBench V6
Key Capabilities
- 超长工具调用链：支持连续 200-300 次工具调用，无需人工干预
- 深度推理能力：在数百个步骤中连贯推理，解决复杂问题
- 测试时扩展：通过扩展思考 token 和工具调用步骤实现测试时扩展
- 多模态工具支持：支持搜索、Python、网页浏览等多种工具
- 开放源代码：完整的权重开源，可通过 Hugging Face 获取
可用性
K2 Thinking 现已在 kimi.com 的聊天模式下上线，完整的智能体模式即将推出。它也可通过 Kimi K2 Thinking API 访问。