Skip to main content
Benchmark Registry
ModelsBenchmarksOrganizations

SWE-bench Pro

Benchmark
SWE-bench Pro
Evaluated by
Scale AI
Release date
September 19, 2025
Version
SWE-bench Pro Public
Metric
Public tasks resolved

Results

3 results

LatestHistory
All providersAnthropicDeepSeek AIGoogle (DeepMind)Meta AI (originally Facebook AI Research)Microsoft AIMiniMaxMoonshot AIOpenAISpaceXAIThinking Machines LabZ.ai
SWE-bench Pro Public results
ProviderModelScoreSourceRegistry No.
OpenAIGPT-5.4 (xhigh)57.7%Source for GPT-5.4 (xhigh) on SWE-bench Pro Public (opens in a new tab)10007
OpenAIGPT-5.3-Codex (xhigh)56.8%Source for GPT-5.3-Codex (xhigh) on SWE-bench Pro Public (opens in a new tab)10006
OpenAIGPT-5.2 (xhigh)55.6%Source for GPT-5.2 (xhigh) on SWE-bench Pro Public (opens in a new tab)10012
Page 1 of 1

© 2026 Densa Labs

Toggle between the data update date and the application build time.
Legal
Color theme