Google DeepMind released its new model Gemini 4 Neon on Nov 18, 2026, alongside a post on its official blog. According to benchmark information circulating on X, Neon is positioned as a model that combines frontier reasoning with fast task completion, aimed primarily at agent workloads. The release sparked discussion on Hacker News, where the related post drew 134 points and 93 comments.

Benchmark performance: Google sweeps LMArena top three

The latest Text Arena rankings published by LMArena show Google DeepMind taking the top three spots: Gemini 4 Carbon (High) leads with 1,526 points, followed by Gemini 4 Argon and Neon in second and third. The result indicates the Gemini 4 series holds a broad lead in text chat evaluations.

Meanwhile, official benchmark tables shared by multiple bloggers on X show Neon beating Claude Opus 5.5 and GPT-6.2 Sol on most of 20 benchmark comparisons. One tally claims Neon leads both rivals on 17 of 20 benchmarks while costing roughly one-third as much. The benchmark tables come from social media and have not been individually confirmed by Google.

LMArena's latest Text Arena rankings show Google DeepMind sweeping the top three: Gemini 4 Carbon (High) leads with 1,526 points, followed by Gemini 4 Argon and Neon in second and third.

Independent evaluation: Artificial Analysis completes four-tier reasoning tests

Independent evaluator Artificial Analysis upgraded its Intelligence Index to v5, adding comparisons with GPT-6.2 and mainstream models and publishing detailed sub-scores covering AA-Briefcase v1.1 (professional knowledge Elo), AA-Omniscience and other items. The organization also completed evaluations of Gemini 4 Neon across four reasoning tiers (Low/Medium/High/Xhigh) and revised its benchmarks.

Pricing strategy and market reaction

According to the circulating benchmark information, Neon is priced at roughly one-third of Claude Opus 5.5 and GPT-6.2 Sol — a significant cost advantage if the figure holds. The heated discussion on Hacker News shows the developer community is paying close attention to the release, though the specific content of the comments was not available in the source material.

Neither Google's official blog nor the DeepMind blog has disclosed detailed pricing or full benchmark data for Neon, and the benchmark tables circulating on social media still need to be checked against official information.

What to watch

  • Whether Google publishes full benchmark data and detailed pricing for Neon
  • The specific results and rankings from Artificial Analysis's four-tier reasoning evaluation
  • Whether Claude Opus 5.5 and GPT-6.2 Sol respond
  • Developer community feedback on Neon's real-world performance in agent workloads