Qwen: Qwen3 30B A3B Thinking 2507 wins on more metrics (6 of 8), but the right pick depends on what you optimize for — see the breakdown below.
| Metric | Mistral: Devstral 2 2512 | Qwen: Qwen3 30B A3B Thinking 2507 |
|---|---|---|
| Intelligence Index | — | — |
| Coding Index | 23.7✓ | 14.6 |
| GPQA Diamond | 59% | 71%✓ |
| Design Arena Elo | — | 973✓ |
| Speed (tokens/sec) | 10 | 161✓ |
| Latency | 897ms | 452ms✓ |
| Input price /M | $0.400 | $0.080✓ |
| Output price /M | $2.00 | $0.400✓ |
| Context window | 262K✓ | 131K |
| Capabilities | ToolsJSON | ReasoningToolsJSON |