← All comparisons
Phi-4 vs Claude 4.5 Sonnet (Reasoning)
Microsoft vs Anthropic — side-by-side benchmark comparison
| Phi-4 | Claude 4.5 Sonnet (Reasoning) | |
|---|---|---|
| Intelligence Index | 10.4 | 43.0 |
| Coding Index | 11.2 | 38.6 |
| Math Index | 18.0 | 88.0 |
| Output speed (tok/s) | 41.1 | 55.0 |
| Blended price ($/1M) | $0.22 | $6.56 |
| Time to first token (s) | 0.50s | 7.02s |
| aime | 14.3% | — |
| aime 25 | 18.0% | 88.0% |
| artificial analysis coding index | 11.20 | 38.60 |
| artificial analysis intelligence index | 10.40 | 43.00 |
| artificial analysis math index | 18.00 | 88.00 |
| gpqa | 57.5% | 83.4% |
| hle | 4.1% | 17.3% |
| ifbench | 23.5% | 57.3% |
| lcr | 0.0% | 65.7% |
| livecodebench | 23.1% | 71.4% |
| math 500 | 81.0% | — |
| mmlu pro | 71.4% | 87.5% |
| scicode | 26.0% | 44.7% |
| tau2 | 0.0% | 78.1% |
| terminalbench hard | 3.8% | 35.6% |
Benchmark data from Artificial Analysis.