Up to four models
DeepSeek-V3.2-Exp vs DeepSeek-V3.1
On the index (GPT-4 = 100), DeepSeek-V3.2-Exp scores 140 and DeepSeek-V3.1 128. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
140124–157
WeightsOpen
ReleasedSep 2025
Price in / out$0.27 / $0.41
Context164K
Core tests taken0 of 9
128111–145
WeightsOpen
ReleasedAug 2025
Price in / out$0.25 / $0.95
Context164K
Core tests taken0 of 9
DeepSeek-V3.2-Exp leads DeepSeek-V3.1 by 12 points, and costs 1.4× less.
Test by testthird-party results
Every result as a table
| Test | DeepSeek-V3.2-Exp | DeepSeek-V3.1 |
|---|---|---|
| Reasoning & knowledge | ||
| Fiction.LiveBench | 83.3% | 52.8% |
| Coding | ||
| WeirdML | 39.5% | 38.4% |
| Agents | ||
| DTBench | 87.7% | 82.7% |
| LMCA | 29.1% | 24.3% |
Tests at least two of these models have taken. Bold rows are the core tests.