← All comparisons
Qwen3 235B A22B 2507 Instruct vs Qwen3 4B 2507 (Reasoning)
Alibaba vs Alibaba — side-by-side benchmark comparison
| Qwen3 235B A22B 2507 Instruct | Qwen3 4B 2507 (Reasoning) | |
|---|---|---|
| Intelligence Index | 25.0 | 18.2 |
| Coding Index | 22.1 | 9.5 |
| Math Index | 71.7 | 82.7 |
| Output speed (tok/s) | 57.0 | 0.0 |
| Blended price ($/1M) | $0.36 | $0.00 |
| Time to first token (s) | 1.34s | 0.00s |
| aime | 71.7% | — |
| aime 25 | 71.7% | 82.7% |
| artificial analysis coding index | 22.10 | 9.50 |
| artificial analysis intelligence index | 25.00 | 18.20 |
| artificial analysis math index | 71.70 | 82.70 |
| gpqa | 75.3% | 66.7% |
| hle | 10.6% | 5.9% |
| ifbench | 46.1% | 49.8% |
| lcr | 31.2% | 37.7% |
| livecodebench | 52.4% | 64.1% |
| math 500 | 98.0% | — |
| mmlu pro | 82.8% | 74.3% |
| scicode | 36.0% | 25.6% |
| tau2 | 33.3% | 25.4% |
| terminalbench hard | 15.2% | 1.5% |
Benchmark data from Artificial Analysis.