GPT-5.6 Luna (Low): benchmark results

GPT-5.6 Luna evaluated at the low reasoning-effort setting. Provider: OpenAI. Released 2026-07-09. Access: API.

Unified ELO 1631 ± 1, rank #313 of 1761 rated models, from 28 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA Omniscience - Software Engineering (SWE)63.8Accuracy (%)86
AA-Omniscience Accuracy39.62Accuracy (%)82.7
AA Omniscience - Business32.4Accuracy (%)82.4
AA Omniscience - Health36Accuracy (%)81.3
AA Omniscience - Science, Engineering & Mathematics41.9Accuracy (%)80.9
AA Omniscience - Humanities & Social Sciences34.9Accuracy (%)79.5
AA Omniscience - Law28.7Accuracy (%)79.3
Artificial Analysis Intelligence Index25.75Intelligence Index73.6
AA GPQA Diamond83.54Accuracy (%)73.5
AA CritPt2.57Accuracy (%)72.1
AA Humanity's Last Exam19.83Accuracy (%)72.1
AA Omniscience-14.65Score68.6

Interactive version: theaggregate.ai/model?slug=gpt-5-6-luna-low · How It Works · Data refreshed daily, snapshot 2026-09-05.