GPT-5.2 (Medium) — benchmark results

GPT-5.2 evaluated at the medium reasoning-effort setting. Provider: OpenAI. Released 2025-12-11. Access: API.

Unified ELO 1899 ± 20, rank #66 of 1776 rated models, from 183 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Elimination Game (Lechmazur)7.52TrueSkill μ100
Epoch AI - Algotune2.05Score100
Medmarks - MTSamples Procedures88.74Score (%)100
Medmarks - MTSamples Replicate98.06Score (%)100
Medmarks - MedCaseReasoning52.6Score (%)100
Medmarks - MedExQA86.93Score (%)100
Medmarks - MedR-Bench 1-Turn84.17Score (%)100
Medmarks - MedR-Bench Free-Turn89.08Score (%)100
Medmarks - MedR-Bench Oracle80.01Score (%)100
Medmarks - MetaMedQA82.88Score (%)100
Medmarks - PubHealthBench Freeform48.09Score (%)100
AA LiveCodeBench89.42Pass@1 (%)99

Interactive version: theaggregate.ai/model?slug=gpt-5-2-medium · How the rankings work · Data refreshed daily, snapshot 2026-07-22.