0 vs 2 benchmarks won
Moonshot AI Kimi K2 | Moonshot AI Kimi K2.5 | |
|---|---|---|
| Overview | ||
| Company | Moonshot AI | Moonshot AI |
| Release date | Jul 11 2025 | Jan 27 2026 |
| Model type | — | — |
| Open source | Yes | Yes |
| Specifications | ||
Parameters | 1T | 1T |
Context window | 128k | 256k |
| Benchmarks | ||
Science reasoning GPQA Diamond | 75.1% | 87.6% |
Software engineering SWE-Bench Verified | 65.8% | 76.8% |
Multimodal understanding MMMU | — | — |
| Timeline | ||
| Release gap | Kimi K2 shipped 200 days before Kimi K2.5 | |
Kimi K2.5 leads Kimi K2 on 2 of the tracked benchmarks (GPQA Diamond, SWE-Bench Verified, MMMU). Kimi K2 shipped 200 days before Kimi K2.5, so benchmark comparisons should account for the intervening progress.
Kimi K2 has 1T parameters, while Kimi K2.5 has 1T. Context windows are 128k (Kimi K2) vs 256k (Kimi K2.5).
On GPQA Diamond, Kimi K2.5 scores 87.6%, 12.5 points above Kimi K2 at 75.1%. On SWE-Bench Verified, Kimi K2.5 scores 76.8%, 11 points above Kimi K2 at 65.8%.