GPT-5.6 Sol (August 2026): benchmark results

Provider: OpenAI. Released 2026-07-09. Access: API.

Unified ELO 1876 ± 39, rank #53 of 1516 rated models, from 45 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
OpenAI GPT-5.6 August Update System Card - Challenging Prompts - Non-Violent Illicit Behavior99.7not_unsafe rate (%)100
OpenAI GPT-5.6 August Update System Card - Challenging Prompts - Self-Harm93.4not_unsafe rate (%)100
OpenAI GPT-5.6 August Update System Card - Challenging Prompts - Violent Illicit Behavior95.9not_unsafe rate (%)100
OpenAI GPT-5.6 August Update System Card - HealthBench55Length-Adjusted Score (0-100)100
OpenAI GPT-5.6 August Update System Card - HealthBench Consensus95.5Length-Adjusted Score (0-100)100
OpenAI GPT-5.6 August Update System Card - HealthBench Hard31.4Length-Adjusted Score (0-100)100
OpenAI GPT-5.6 August Update System Card - HealthBench Professional54Length-Adjusted Score (0-100)100
OpenAI GPT-5.6 August Update System Card - Image Input - Self-Harm92.3not_unsafe rate (%)100
OpenAI GPT-5.6 August Update System Card - U18 - Age-Restricted Goods and Dangerous Activities86.5Score (%)100
OpenAI GPT-5.6 August Update System Card - U18 - Gore87.2Score (%)100
OpenAI GPT-5.6 August Update System Card - U18 - Self-Harm98.7Score (%)100
OpenAI MentalHealthBench - Acuity - High Acuity53.59Task-clipped rubric score (%)81.2

Interactive version: theaggregate.ai/model?slug=gpt-5-6-sol-august-2026 · How It Works · Data refreshed daily, snapshot 2026-09-24.