Gemini 4 Neon Benchmark Sheet: Beats Opus 5.5 and Sol 6.2 on 17/20, at One-Third the Cost

imjustnewatai · x · 2026-11-19

A 20-benchmark comparison sheet for Gemini 4 Neon shows it beating Opus 5.5 and Sol 6.2 on 17 of 20, and both Astra 6.2 and Fable 5.2 on 4 of 12 where they have published scores — near-frontier performance at roughly one-third the cost. Highlights: top AA Coding Agent Index (58.6), SWE-sweep 39.0, AutomationBench 56.3, SEC-bench Pro 90.5; but overall AA score 61 trails Astra 6.2's 62, and HLE-Diamond 71.3 lags Fable 5.2's 80.2. Note: mostly vendor-reported scores with non-identical setups.

Related event: Gemini 4 Neon benchmark tables leak showing leads over Opus 5.5 and GPT-6.2 Sol(5 posts)→

Original post →

More from Models

Models channel →