GPT 5.4 Pro 2026 03.05 Web App

released 2026-03-05

BenchAtlas Index

as of 2026-09-11

Not enough benchmark coverage for a composite score (needs ≥4 families across ≥3 categories). The per-benchmark evidence below stands on its own — sparse coverage is an evidence gap, not a low score.

Benchmark evidence

2 results

external indices

ECI

158.9 points

independent

Epoch AI Benchmarking Hub

reasoning math

FrontierMath Tier 4older version

37.5 %

independent

Epoch AI Benchmarking Hub

2026-03-06