Google Neon Release - SWE-sweep - mini-swe vision: leaderboard

Google Gemini 4 Neon launch evaluation on SWE-sweep software engineering tasks using the mini-swe vision harness; subtask-weighted reward (%).

Metric: Subtask-weighted reward (%). Source: storage.googleapis.com. 1 model tracked. Updated 19 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 4 NeonxHigh12.8#3

Interactive version: theaggregate.ai/benchmark?slug=google-neon-release-swe-sweep-mini-swe-vision · How It Works · Data refreshed daily, snapshot 2026-11-20.