DeepSeek: DeepSeek V3.2 Exp wins on more metrics (5 of 9), but the right pick depends on what you optimize for — see the breakdown below.
| Metric | Anthropic: Claude Sonnet 4 | DeepSeek: DeepSeek V3.2 Exp |
|---|---|---|
| Intelligence Index | 33.0✓ | 32.1 |
| Coding Index | 30.6 | 34.6✓ |
| GPQA Diamond | 68% | 75%✓ |
| Design Arena Elo | 1220 | 1229✓ |
| Speed (tokens/sec) | 50✓ | 20 |
| Latency | 720ms✓ | 1.3s |
| Input price /M | $3.00 | $0.270✓ |
| Output price /M | $15.00 | $0.410✓ |
| Context window | 1M✓ | 164K |
| Capabilities | ReasoningToolsVision | ReasoningToolsJSON |