GPT-6 Luna: benchmark results
Intended for focused tasks repeated at volume. Supports structured outputs and tool calls, with reasoning that can be switched off for simpler requests. Provider: OpenAI. Released 2026-09-22. Access: API.
Aggregate Rating 1820 ± 7, rank #52 of 1635 rated models, from 493 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA-AnalystAgent - Pass@K | 85 | Pass@K (%) | 100 |
| OpenAI GPT-6 Astra System Card - Agentic Safe Completions - Codex Non-Violent Wrongdoing | 100 | Safe completion rate (%) | 100 |
| OpenAI GPT-6 Astra System Card - Agentic Safe Completions - Codex Self-Harm | 96 | Safe completion rate (%) | 100 |
| OpenAI GPT-6 Astra System Card - Image Input - Self-Harm | 99.9 | Safe completion rate (%) | 100 |
| OpenAI MentalHealthBench - Behavior - Interpretation | 79.62 | Axis score, normalized to its possible range (%) | 100 |
| Icelandic LLM - Generative Quality | 82.26 | Quality Score (%) | 97.5 |
| AA Long Context Reasoning | 83.33 | Accuracy (%) | 96.3 |
| MLCR-AA - Accuracy | 93.75 | Accuracy (%) | 95.8 |
| PLCC - Grammar | 93 | Accuracy (%) | 94.5 |
| PLCC - History | 96 | Accuracy (%) | 94.5 |
| Coarena - Task Completion | 80.3 | Completion Rate (%) | 94.1 |
| SvelteBench | 100 | Average pass@1 (%) | 94 |
Interactive version: theaggregate.ai/model?slug=gpt-6-luna · How It Works · Data refreshed daily, snapshot 2026-10-15.