OpenAI: gpt-oss-20b wins on more metrics (6 of 9), but the right pick depends on what you optimize for — see the breakdown below.
| Metric | MoonshotAI: Kimi K2 Thinking | OpenAI: gpt-oss-20b |
|---|---|---|
| Intelligence Index | 24.1 | 24.5✓ |
| Coding Index | 15.5 | 18.5✓ |
| GPQA Diamond | 71%✓ | 69% |
| Design Arena Elo | 1158✓ | 977 |
| Speed (tokens/sec) | 103 | 397✓ |
| Latency | 412ms | 223ms✓ |
| Input price /M | $0.600 | $0.029✓ |
| Output price /M | $2.50 | $0.140✓ |
| Context window | 262K✓ | 131K |
| Capabilities | ReasoningToolsJSON | ReasoningToolsJSON |