← All comparisons
DeepSeek R1 Distill Llama 70B vs Qwen3 VL 235B A22B (Reasoning)
DeepSeek vs Alibaba — side-by-side benchmark comparison
| DeepSeek R1 Distill Llama 70B | Qwen3 VL 235B A22B (Reasoning) | |
|---|---|---|
| Intelligence Index | 16.0 | 27.6 |
| Coding Index | 11.4 | 20.9 |
| Math Index | 53.7 | 88.3 |
| Output speed (tok/s) | 46.8 | 35.6 |
| Blended price ($/1M) | $0.79 | $2.17 |
| Time to first token (s) | 0.33s | 5.14s |
| aime | 67.0% | — |
| aime 25 | 53.7% | 88.3% |
| artificial analysis coding index | 11.40 | 20.90 |
| artificial analysis intelligence index | 16.00 | 27.60 |
| artificial analysis math index | 53.70 | 88.30 |
| gpqa | 40.2% | 77.2% |
| hle | 6.1% | 10.1% |
| ifbench | 27.6% | 56.5% |
| lcr | 11.0% | 58.7% |
| livecodebench | 26.6% | 64.6% |
| math 500 | 93.5% | — |
| mmlu pro | 79.5% | 83.6% |
| scicode | 31.3% | 39.9% |
| tau2 | 21.9% | 54.1% |
| terminalbench hard | 1.5% | 11.4% |
Benchmark data from Artificial Analysis.