← All comparisons
Claude 4.5 Sonnet (Reasoning) vs DeepSeek V3.1 Terminus (Reasoning)
Anthropic vs DeepSeek — side-by-side benchmark comparison
| Claude 4.5 Sonnet (Reasoning) | DeepSeek V3.1 Terminus (Reasoning) | |
|---|---|---|
| Intelligence Index | 43.0 | 33.9 |
| Coding Index | 38.6 | 33.7 |
| Math Index | 88.0 | 89.7 |
| Output speed (tok/s) | 55.0 | 0.0 |
| Blended price ($/1M) | $6.56 | $1.91 |
| Time to first token (s) | 7.02s | 0.00s |
| aime | — | — |
| aime 25 | 88.0% | 89.7% |
| artificial analysis coding index | 38.60 | 33.70 |
| artificial analysis intelligence index | 43.00 | 33.90 |
| artificial analysis math index | 88.00 | 89.70 |
| gpqa | 83.4% | 79.2% |
| hle | 17.3% | 15.2% |
| ifbench | 57.3% | 57.0% |
| lcr | 65.7% | 65.0% |
| livecodebench | 71.4% | 79.8% |
| math 500 | — | — |
| mmlu pro | 87.5% | 85.1% |
| scicode | 44.7% | 40.6% |
| tau2 | 78.1% | 37.1% |
| terminalbench hard | 35.6% | 30.3% |
Benchmark data from Artificial Analysis.