GPT-OSS-120B (High): benchmark results
GPT-OSS-120B evaluated at the high reasoning-effort setting. Provider: OpenAI. Released 2025-08-05. Access: Open.
Aggregate Rating 1607 ± 6, rank #690 of 2168 rated models, from 487 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ESF-Bench - Long and Multi-Value | 81.5 | Tagged slot accuracy (%; slots tagged with a LMV (Long and M | 100 |
| ESF-Bench - Schema Constraints | 83.8 | Tagged slot accuracy (%; slots tagged with a SC (Schema Cons | 100 |
| LLM Stats (Codeforces Rating - GPT-OSS no tools) | 2463 | Codeforces Rating | 100 |
| MolecularIQ - Counting | 46.8 | Mean Accuracy (%) | 100 |
| MolecularIQ - Generation | 53.7 | Mean Accuracy (%) | 100 |
| MolecularIQ - Indexing | 42.5 | Mean Accuracy (%) | 100 |
| MolecularIQ - Overall | 47.5 | Mean Accuracy (%) | 100 |
| NoisyBench - Hard Negative Distractors | 50.2 | Harmonic Mean Score (%; 11 datasets' pass@k, pass^k for TauB | 100 |
| OdysseyArena (Rules Given) - AI Trading | 141.63 | Profit rate in percent averaged over four runs per task (Avg | 100 |
| OpenCompass Code - Competition - English (CompassBench 2510) | 75.4 | Score (%) | 100 |
| OpenCompass LLM - Code (CompassBench 2510) | 72.5 | Score (%) | 100 |
| OpenCompass LLM - Language (CompassBench 2510) | 89.3 | Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-oss-120b-high · How It Works · Data refreshed daily, snapshot 2026-10-20.