Google docs reveal Gemini 4 lineup: Carbon tops reasoning, Hydrogen cuts hallucinations 5x vs Flash
cantrell · x · 2026-11-20
A leaked Google model-selection doc confirms the Gemini 4 lineup: Carbon is the strongest reasoning model (system architecture, multi-file debugging, proofs, security audits); Neon sits between Hydrogen and Carbon with deeper reasoning than Hydrogen but fewer output tokens than Carbon (paid tier only); Hydrogen is the new default everyday model, cutting fabricated claims by more than 5x vs Gemini 3.8 Flash by flagging missing context, with parallel read/search tool calls. Gemini 3.5 Flash-Lite remains the cheapest high-volume option, alongside specialist models like gemini-decision-1, Nano Banana, Gemini Omni and Gemini 3.8 Live. Author cantrell notes Hydrogen isn't a direct Flash replacement, Argon appears replaced by Neon, and criticizes Google's confusing element-based naming scheme.
Related event: Gemini 4 Naming Chaos Confuses Users as Family Docs Leak(4 posts)→
More from Models
- Naval-war game benchmark: author claims GPT 6.2 Astra crushes Opus 5.2 — RichardsonDx · 2026-11-20
- Qwen3.8-27B-pi-adapted GGUF quantization trends on Hugging Face — harness-hub · 2026-11-20
- One neuron per language hits 99.1% language-ID accuracy, beating dedicated models — julianweisser · 2026-11-20
- 10 hands-on lessons on what actually happens inside an AI after you type a prompt — techartist_ · 2026-11-20
- Redwood questions Google reusing safety results from weaker model for stronger Carbon — YvesMulkers · 2026-11-20
- GTE-MLM sparse embedding models now natively supported in Transformers v5 — tomaarsen · 2026-11-20