← All comparisons
Gemini 2.5 Pro vs Claude 4.5 Haiku (Non-reasoning)
Google vs Anthropic — side-by-side benchmark comparison
| Gemini 2.5 Pro | Claude 4.5 Haiku (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 34.6 | 31.0 |
| Coding Index | 32.0 | 29.6 |
| Math Index | 87.7 | 39.0 |
| Output speed (tok/s) | 140.4 | 105.8 |
| Blended price ($/1M) | $3.44 | $2.19 |
| Time to first token (s) | 20.87s | 0.65s |
| aime | 88.7% | — |
| aime 25 | 87.7% | 39.0% |
| artificial analysis coding index | 32.00 | 29.60 |
| artificial analysis intelligence index | 34.60 | 31.00 |
| artificial analysis math index | 87.70 | 39.00 |
| gpqa | 84.4% | 64.6% |
| hle | 21.1% | 4.3% |
| ifbench | 48.7% | 42.0% |
| lcr | 66.0% | 43.7% |
| livecodebench | 80.1% | 51.1% |
| math 500 | 96.7% | — |
| mmlu pro | 86.2% | 80.0% |
| scicode | 42.8% | 34.4% |
| tau2 | 54.1% | 32.5% |
| terminalbench hard | 26.5% | 27.3% |
Benchmark data from Artificial Analysis.