GPT-5.6 Terra (Medium): benchmark results

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

Unified ELO 1674 ± 1, rank #168 of 1761 rated models, from 39 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA CritPt17.43Accuracy (%)90.5
AA Omniscience - Software Engineering (SWE)69.6Accuracy (%)90.5
AA Omniscience - Health42.2Accuracy (%)89.9
Artificial Analysis Intelligence Index37.23Intelligence Index88.9
AA Omniscience - Humanities & Social Sciences41.6Accuracy (%)87.7
LLM Chess (Saplin)935.9ELO87.6
AA-Omniscience Accuracy44.6Accuracy (%)87.5
LisanBench0.21Mean Path Length / Current Maximum85.6
AA Omniscience - Science, Engineering & Mathematics44.4Accuracy (%)85.4
AA Omniscience - Law36.3Accuracy (%)85.2
AA Humanity's Last Exam33.27Accuracy (%)84.2
AA Omniscience - Business33.5Accuracy (%)83.4

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