← All comparisons
DeepSeek R1 Distill Qwen 14B vs Kimi K2.5 (Non-reasoning)
DeepSeek vs Kimi — side-by-side benchmark comparison
| DeepSeek R1 Distill Qwen 14B | Kimi K2.5 (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 15.8 | 37.3 |
| Coding Index | — | 25.8 |
| Math Index | 55.7 | — |
| Output speed (tok/s) | 0.0 | 33.5 |
| Blended price ($/1M) | $0.00 | $1.20 |
| Time to first token (s) | 0.00s | 1.23s |
| aime | 66.7% | — |
| aime 25 | 55.7% | — |
| artificial analysis coding index | — | 25.80 |
| artificial analysis intelligence index | 15.80 | 37.30 |
| artificial analysis math index | 55.70 | — |
| gpqa | 48.4% | 78.9% |
| hle | 4.4% | 12.3% |
| ifbench | 22.1% | 43.7% |
| lcr | 7.0% | 59.0% |
| livecodebench | 37.6% | — |
| math 500 | 94.9% | — |
| mmlu pro | 74.0% | — |
| scicode | 23.9% | 39.6% |
| tau2 | — | 81.3% |
| terminalbench hard | — | 18.9% |
Benchmark data from Artificial Analysis.