Kimi K2 Thinking
- Made by
- Moonshot AI
- Released
- Announced in
- Introducing Kimi K2 Thinkingkimi.ai
- Event
- Kimi K2 Thinking
- API record
- /api/v1/models/kimi-k2-thinking
Benchmark results
| Benchmark | Score | Source check |
|---|---|---|
| Humanity's Last ExamNo tools · text-only subset; 96K thinking-token budget; INT4 precision | 23.9% | CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Humanity's Last ExamWith tools · text-only subset; search, code and browsing tools; up to 120 steps; Hugging Face blocked; INT4 precision | 44.9% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Hugging Face access not blocked, which may leak answers; text-only subset | 51.3% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| heavy mode: eight parallel trajectories aggregated into one answer; text-only subset · parallel test-time compute | 51.0% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| GPQA Diamond96K thinking-token budget; INT4 precision | 84.5% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| BrowseCompsearch, code and browsing tools; up to 300 steps; older tool outputs hidden past 256K tokens | 60.2% | CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
Sources
Reported by Moonshot AI
- Introducing Kimi K2 Thinkingkimi.ai · 6 Nov 2025
Our capture of the source