← All comparisons
Qwen3.5 397B A17B (Non-reasoning) vs DeepSeek V3.1 Terminus (Non-reasoning)
Alibaba vs DeepSeek — side-by-side benchmark comparison
| Qwen3.5 397B A17B (Non-reasoning) | DeepSeek V3.1 Terminus (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 40.1 | 28.5 |
| Coding Index | 37.4 | 31.9 |
| Math Index | — | 53.7 |
| Output speed (tok/s) | 53.5 | 0.0 |
| Blended price ($/1M) | $1.35 | $0.45 |
| Time to first token (s) | 1.85s | 0.00s |
| aime | — | — |
| aime 25 | — | 53.7% |
| artificial analysis coding index | 37.40 | 31.90 |
| artificial analysis intelligence index | 40.10 | 28.50 |
| artificial analysis math index | — | 53.70 |
| gpqa | 86.1% | 75.1% |
| hle | 18.8% | 8.4% |
| ifbench | 51.6% | 41.2% |
| lcr | 58.0% | 43.3% |
| livecodebench | — | 52.9% |
| math 500 | — | — |
| mmlu pro | — | 83.6% |
| scicode | 41.1% | 32.1% |
| tau2 | 83.9% | 37.1% |
| terminalbench hard | 35.6% | 31.8% |
Benchmark data from Artificial Analysis.