GPT-5 Nano (Medium): benchmark results

GPT-5 Nano evaluated at the medium reasoning-effort setting. Provider: OpenAI. Released 2025-08-07. Access: API.

Unified ELO 1528 ± 1, rank #725 of 1761 rated models, from 75 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
MATH Level 595.24Accuracy (%)87
AA LiveCodeBench76.3Pass@1 (%)83.1
AA IFBench65.92Accuracy (%)77.4
LisanBench0.08Mean Path Length / Current Maximum71.9
AA AIME 202578.33Accuracy (%)71.1
LLM Chess (Saplin)413.3ELO70.8
AA Omniscience - Software Engineering (SWE) - JavaScript33.64Accuracy (%)69.1
AI for Education Visual Maths - Algebra84.62Accuracy (%)64.9
AA Omniscience - Software Engineering (SWE) - Kotlin20Accuracy (%)63
SLR-Bench - Basic99Accuracy (%)62.7
AI for Education Visual Maths - Measurement83.78Accuracy (%)60.8
AI for Education Visual Maths - Statistics and Probability28.57Accuracy (%)60.8

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