GPT-5.5 (Low): benchmark results
GPT-5.5 evaluated at the low reasoning-effort setting. Provider: OpenAI. Released 2026-04-23. Access: API.
Aggregate Rating 1779 ± 15, rank #174 of 2168 rated models, from 92 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| K-Bench | 98.96 | Overall (%, Default v1 prompt) | 100 |
| MAG (Coordinates) | 37.4 | Task success rate (%; functional WebArena checkers, LLM judg | 100 |
| MAG (Coordinates) - Gated Guide Score | 22.5 | Gated guide score (%): zero on a failed task and 0.4 + 0.6 t | 100 |
| MAG (Set-of-Mark) | 35.6 | Task success rate (%; functional WebArena checkers, LLM judg | 100 |
| MAG (Set-of-Mark) - Gated Guide Score | 20.6 | Gated guide score (%): zero on a failed task and 0.4 + 0.6 t | 100 |
| K-Bench (Therapeutic v0) | 98.72 | Overall (%, Therapeutic v0 prompt) | 98.4 |
| AA-Omniscience Index - Software Engineering (SWE) - PHP | 84 | Omniscience Index | 97.6 |
| PLCC - Geography | 99 | Accuracy (%) | 97.1 |
| K-Bench - Risk | 95.51 | Risk Score (%, Default v1 prompt) | 96.7 |
| PLCC - Art & Entertainment | 93 | Accuracy (%) | 95.9 |
| Sonar LLM Leaderboard - Java - Bug Density | 0.55 | Bugs per 1,000 lines of code (lower is better) | 95.8 |
| AA Terminal-Bench Hard | 52.27 | Accuracy (%) | 95.7 |
Interactive version: theaggregate.ai/model?slug=gpt-5-5-low · How It Works · Data refreshed daily, snapshot 2026-10-20.