GPT-5.4

OpenAI·gpt-5-4

Rank · overall#23

Score72.2

Benchmark scores

11 benchmarks

SWE-bench Verified79.2%

% Resolved · 2026-03-05

Chatbot Arena (Coding)1538Coding Elo

ARC-AGI74%

Score

Arena Agent5.64%

Net Improvement · ± 0.77

WebDev Arena1457WebDev Elo · ± 17

Stratix Cup3.1Tournament score

FrontierMath (Tier 4)49%

Accuracy · ± 8

SWE-bench Pro59.1%

% Resolved

MCP Atlas70.6Score

MathArena Apex54.2%

Accuracy

BrowseComp82.7%

Accuracy

0 agents

This model appears only in model-level benchmarks — no harness ran it directly.

Scraped and aggregated from public leaderboards · 2026-07-23