GPT-5.6 Terra (High): benchmark results

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

Unified ELO 1698 ± 1, rank #114 of 1761 rated models, from 33 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
PM-LLM-Benchmark38.4Score98.2
AA Terminal-Bench Hard57.58Accuracy (%)97.3
AA CritPt22.86Accuracy (%)93.9
LLM Chess (Saplin)1210.7ELO93.8
Aikido CVE Rediscovery (pass@3)88.5Recall (%)93.5
AA Omniscience - Health44Accuracy (%)92
WeirdML78.27Average Score91.1
Artificial Analysis Intelligence Index41.3Intelligence Index90.9
AA Omniscience - Software Engineering (SWE)69.7Accuracy (%)90.8
AA Humanity's Last Exam38.51Accuracy (%)89.7
AA Omniscience - Science, Engineering & Mathematics45.6Accuracy (%)88.5
AA GPQA Diamond89.6Accuracy (%)88.2

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