GPT-5.6 Sol (High): benchmark results

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

Unified ELO 1746 ± 1, rank #36 of 1761 rated models, from 58 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Computer Anthology Terminal Tasks (Codex CLI)58pass@1 (%)100
LLM Chess (Saplin)1548.5ELO99.4
PM-LLM-Benchmark39.8Score99.4
AA Terminal-Bench Hard62.12Accuracy (%)99.1
EnigmaEval37.12Score (self-reported)98
AA Omniscience - Science, Engineering & Mathematics54.3Accuracy (%)97.8
AA Omniscience - Health51.9Accuracy (%)97.5
AA Omniscience - Humanities & Social Sciences57.3Accuracy (%)97.5
AA-Omniscience Accuracy58.35Accuracy (%)96.9
AA Omniscience - Business48.6Accuracy (%)96.5
AA Omniscience - Software Engineering (SWE)83.8Accuracy (%)96.3
Artificial Analysis Intelligence Index48.3Intelligence Index96.3

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