1 vs 1 benchmarks won
Google Gemini 2.5 Pro | Moonshot AI Kimi K2 (0905) | |
|---|---|---|
| Overview | ||
| Company | Moonshot AI | |
| Release date | Mar 25 2025 | Sep 5 2025 |
| Model type | — | — |
| Open source | No | Yes |
| Specifications | ||
Parameters | — | 1T |
Context window | — | 256k |
| Benchmarks | ||
Coding SWE-Bench VerifiedReal coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. | 59.6% | 65.8%Best |
Science GPQA DiamondGraduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 86.4%Best | 75.1% |
Multimodal MMMUTests the AI on understanding images and text together across many college subjects. Higher is better. | 68% | — |
| Timeline | ||
| Release gap | Gemini 2.5 Pro shipped 164 days before Kimi K2 (0905) | |
Gemini 2.5 Pro and Kimi K2 (0905) are evenly matched across the 2 benchmarks they both report (SWE-Bench Verified, GPQA Diamond). Gemini 2.5 Pro shipped 164 days before Kimi K2 (0905), so benchmark comparisons should account for the intervening progress.
Kimi K2 (0905) is an open-source / open-weight model; Gemini 2.5 Pro is proprietary.
On SWE-Bench Verified, Kimi K2 (0905) leads at 65.8% vs Gemini 2.5 Pro at 59.6%. On GPQA Diamond, Gemini 2.5 Pro leads at 86.4% vs Kimi K2 (0905) at 75.1%.