Google Reportedly Quietly Released Gemini 4 Neon, Topping AA Index at 61
heypearlai · x · 2026-11-22
An account claims Google quietly shipped Gemini 4 Neon amid the Claude/GPT face-off (unverified by Google) and published an official-style comparison vs Claude Opus 5.5 and GPT-6.2 Sol:
- Overall: tops AA Index at 61 (Opus 59, GPT-6.2 57); Vals Index 70.9%
- Agentic coding: CursorBench 4.1 at 65.4% (Opus 58.1%); SWE-sweep provider harness 39.0% — but loses Terminal-Bench 4.1 (65.2%) and FrontierCode 2.0 (46.2%) to Opus 5.5
- Science & math: RiemannBench 84.0%, HLE-Diamond 71.3%, both first
- Long context: GraphWalks (256k–1M) 95.2% vs Opus 66.8%
- Security: SEC-bench Pro 90.5%, WebExploitBench 51.3%
The bigger claim: what looked like a two-horse OpenAI–Anthropic race six months ago is now a wide-open field with Google back and Meta pushing hard.
More from Models
- Forecasters move superintelligence timelines up after METR finds AI R&D speed doubled — ArtificialOther · 2026-11-22
- AI model bizarrely references an image the user never uploaded — lyraxana · 2026-11-22
- Blogger claims Gemini 4 Neon one-shots bug-free iOS apps, outclasses GPT-6.2 and Claude Opus 5.5 — kimmonismus · 2026-11-22
- Bindu Reddy predicts OpenAI will ship Astra 6.3 next week and reclaim the top spot — bindureddy · 2026-11-22
- Boaz Barak says ML theorists already find OpenAI's internal Astra model productive — boazbaraktcs · 2026-11-22
- Power user's verdict: Gemini Neon plans best, Claude still builds, Kimi held back by hardware — Quick-Collection6281 · 2026-11-22