← All comparisons
Llama 3.1 Instruct 405B vs Qwen3 235B A22B (Reasoning)
Meta vs Alibaba — side-by-side benchmark comparison
| Llama 3.1 Instruct 405B | Qwen3 235B A22B (Reasoning) | |
|---|---|---|
| Intelligence Index | 17.4 | 19.8 |
| Coding Index | 14.5 | 17.4 |
| Math Index | 3.0 | 82.0 |
| Output speed (tok/s) | 37.5 | 58.3 |
| Blended price ($/1M) | $3.69 | $2.63 |
| Time to first token (s) | 0.63s | 1.37s |
| aime | 21.3% | 84.0% |
| aime 25 | 3.0% | 82.0% |
| artificial analysis coding index | 14.50 | 17.40 |
| artificial analysis intelligence index | 17.40 | 19.80 |
| artificial analysis math index | 3.00 | 82.00 |
| gpqa | 51.5% | 70.0% |
| hle | 4.2% | 11.7% |
| ifbench | 39.0% | 38.7% |
| lcr | 24.3% | 0.0% |
| livecodebench | 30.5% | 62.2% |
| math 500 | 70.3% | 93.0% |
| mmlu pro | 73.2% | 82.8% |
| scicode | 29.9% | 39.9% |
| tau2 | 19.0% | 24.0% |
| terminalbench hard | 6.8% | 6.1% |
Benchmark data from Artificial Analysis.