Qwen3 235B-A22B Instruct 2507 scores 18.5 on the Respan Index (#101), ahead of GPT-5 mini at 18.3 (#103). They split the 2 benchmarks both report evenly. Qwen3 235B-A22B Instruct 2507 is 1.7x cheaper per token at list price. GPT-5 mini has the larger context window (400K tokens).
| Benchmark | GPT-5 mini | Qwen3 235B-A22B Instruct 2507 |
|---|---|---|
| WeirdML (Håvard Tveit Ihle) | 52.7% | 38.7% |
| LMArena Elo | 1390 | 1422 |
Higher is better. Each score links to where it was published.
* Estimated: no published scores in that category. How the Respan Index works