DeepSeek V3.2 Exp (Non-reasoning) — benchmark results
DeepSeek V3.2 Exp evaluated with reasoning disabled. Provider: DeepSeek. Released 2025-09-29. Access: Open.
Unified ELO 1637 ± 18, rank #396 of 1776 rated models, from 39 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| UGI Leaderboard | 54.67 | UGI Score | 96.2 |
| UGI - Natural Intelligence | 47.5 | NatInt Score | 89.9 |
| UGI - Writing | 51.04 | Writing Score | 89.3 |
| AA MMLU-Pro | 83.64 | Accuracy (%) | 85.5 |
| AA CritPt | 1.43 | Accuracy (%) | 75.1 |
| AA Omniscience - Software Engineering (SWE) - Go | 30 | Accuracy (%) | 74.3 |
| AA SciCode | 39.93 | Accuracy (%) | 73.1 |
| AA Omniscience - Humanities & Social Sciences | 26 | Accuracy (%) | 70.3 |
| AA Omniscience - Software Engineering (SWE) - Python | 30 | Accuracy (%) | 69.4 |
| AA Omniscience - Software Engineering (SWE) - Kotlin | 26 | Accuracy (%) | 68.5 |
| AA Omniscience - Software Engineering (SWE) - R | 20 | Accuracy (%) | 66.6 |
| AA Omniscience - Law | 15.7 | Accuracy (%) | 66.4 |
Interactive version: theaggregate.ai/model?slug=deepseek-v3-2-exp-non-reasoning · How the rankings work · Data refreshed daily, snapshot 2026-07-22.