Coding / Agent
FreshHigh confidenceSWE-benchSWE-bench bash-only · result 2026-02-26Updated: 8/10/2026
69.6% resolved
#15 · 70th percentile
3 Pro
Aliases: gemini-3-pro, gemini-3-pro-preview
google-gemini-3-proSource ranks stay intact; incompatible raw scores are not added.
This model has scores from 2 independent sources and is eligible for the ToolCenter composite reference.
Historical snapshots are available on the source score cards.