Gemini 4 Neon full scorecard: beats Carbon on 15 of 17 benchmarks, edges Opus 5.5 on most evals
lyraxana · x · 2026-11-19
A community-compiled official scorecard for Gemini 4 Neon, with Gemini 4 Hydrogen, Carbon, and Argon columns added from public leaderboards and model cards (as of Nov 18):
- Neon beats in-family Carbon on 15 of 17 shared benchmarks
- Versus Claude Opus 5.5 and GPT-6.2 Sol, Neon leads on nearly every reported eval, losing narrowly only Terminal-bench 4.1 (65.2% vs 65.8%) and FrontierCode 2.0 (46.2% vs 47.3%) to Opus 5.5, plus Chartography to Sol
- Standouts: GraphWalks long-context 95.2%, SEC-bench Pro 90.5%, Agents' Last Exam 75.8%, HLE-Diamond 71.3%
- Methodology footnotes flag which numbers are official vs self-compiled
Caveat: "leads on everything" deserves discounting—Opus 5.5 still edges coding benchmarks, and several figures come from different harnesses.
More from Models
- Cloudflare open-weights Clairic: 27B and 9B models built for extraction — michellechen · 2026-11-19
- Gemini Neon test: it names which "AI SEO courses" are rebranded legacy content and which to skip — gaganghotra_ · 2026-11-19
- Gemini 4 Neon's 39% SWE-sweep questioned: results hinge on the harness — OfirPress · 2026-11-19
- scaling01 mocks benchmark tables: leading every eval doesn't mean the better model — scaling01 · 2026-11-19
- Microsoft launches MAI Realtime 1: speech-to-speech voice AI at $0.01 per audio minute — alexcovo_eth · 2026-11-19
- Gemini 4 Neon appears live and quick: researcher shares one-shot demos, best scores 7/10 — SamuelAlbanie · 2026-11-19