0 vs 2 benchmarks won
Anthropic Claude Haiku 4.5 | Moonshot AI Kimi K2.5 | |
|---|---|---|
| Overview | ||
| Company | Anthropic | Moonshot AI |
| Release date | Oct 15 2025 | Jan 27 2026 |
| Model type | — | — |
| Open source | No | Yes |
| Specifications | ||
Parameters | — | 1T |
Context window | — | 256k |
| Benchmarks | ||
Science reasoning GPQA Diamond | 73% | 87.6% |
Software engineering SWE-Bench Verified | 73.3% | 76.8% |
Multimodal understanding MMMU | — | — |
| Timeline | ||
| Release gap | Claude Haiku 4.5 shipped 104 days before Kimi K2.5 | |
Kimi K2.5 leads Claude Haiku 4.5 on 2 of the tracked benchmarks (GPQA Diamond, SWE-Bench Verified, MMMU). Claude Haiku 4.5 shipped 104 days before Kimi K2.5, so benchmark comparisons should account for the intervening progress.
Kimi K2.5 is an open-source / open-weight model; Claude Haiku 4.5 is proprietary.
On GPQA Diamond, Kimi K2.5 scores 87.6%, 14.6 points above Claude Haiku 4.5 at 73%. On SWE-Bench Verified, Kimi K2.5 scores 76.8%, 3.5 points above Claude Haiku 4.5 at 73.3%.