← All comparisons
Claude 4.1 Opus (Reasoning) vs DeepSeek V3.2 (Non-reasoning)
Anthropic vs DeepSeek — side-by-side benchmark comparison
| Claude 4.1 Opus (Reasoning) | DeepSeek V3.2 (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 42.0 | 32.1 |
| Coding Index | 36.5 | 34.6 |
| Math Index | 80.3 | 59.0 |
| Output speed (tok/s) | 44.5 | 0.0 |
| Blended price ($/1M) | $32.81 | $0.78 |
| Time to first token (s) | 8.55s | 0.00s |
| aime | — | — |
| aime 25 | 80.3% | 59.0% |
| artificial analysis coding index | 36.50 | 34.60 |
| artificial analysis intelligence index | 42.00 | 32.10 |
| artificial analysis math index | 80.30 | 59.00 |
| gpqa | 80.9% | 75.1% |
| hle | 11.9% | 10.5% |
| ifbench | 55.4% | 49.0% |
| lcr | 66.3% | 39.0% |
| livecodebench | 65.4% | 59.3% |
| math 500 | — | — |
| mmlu pro | 88.0% | 83.7% |
| scicode | 40.9% | 38.7% |
| tau2 | 71.4% | 78.9% |
| terminalbench hard | 34.3% | 32.6% |
Benchmark data from Artificial Analysis.