Google releases Gemini 4 Neon
Priced at $3 and $15 per million tokens with a 1M-token window; the model card says its gain over Argon fell below Google's threshold for a full critical-capability evaluation.
- Models & capabilities
- Notable
Google released Gemini 4 Neon, the fourth model in the Gemini 4 line after Argon, Carbon and Hydrogen. Google describes it as balancing depth and efficiency between Carbon and Hydrogen, aimed at agentic coding, knowledge work and science. It is priced at $3 per million input tokens, $0.15 for cached input and $15 per million output tokens, with a 1 million-token context window. It is available in the Gemini API and Google AI Studio, in Gemini for business, in Antigravity and the Gemini app for paid Pro and Ultra subscribers, with a cyber-defence variant promised to vetted defenders in the Fairwind Program “in the coming days”.
The model card says Gemini 4 Neon is adapted from Gemini 4 Carbon, with architecture, data and hardware described by reference to Carbon’s card. Its knowledge cutoff is March 2026.
Nearly every performance claim in the launch post is Google’s own, citing third-party boards. Google said Neon scores 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, “#3 overall”, ranks second on the Vals Index (70.9%), and is “best in class” on Terminal-Bench Science (70.5%), RiemannBench (84.0%) and HLE-Diamond (71.3%). It also claimed a 32 per cent lower cost per task than Carbon at the default high setting. Artificial Analysis’s public leaderboard, read on 22 November, lists Neon at 61, behind Claude Fable 5.2 at 65 and GPT-6.2 Astra at 62 (see OpenAI’s release), and ahead of Carbon at 60.
The more consequential disclosure is in the model card. Under its Frontier Safety Framework, Google ran a Material Capability Change assessment against Gemini 4 Argon, the latest model to have had a full critical-capability evaluation. Neon showed a general capability increase, but below the threshold, so the full evaluations were not run. Google carried Argon’s determinations forward: the CBRN uplift threshold was not reached, but the biological-weapons and cybersecurity critical capability levels “cannot be ruled out”, with mitigations deployed. A full Safety and Security Report was promised “in the coming days”. In effect, Neon’s safety determinations rest on Argon’s evaluation rather than a fresh one, which Google’s framework permits below the threshold.
The release continues Google’s tiered approach to the Gemini 4 family, and puts a model priced below the top tier close to the leaders on at least one independent index.
In the commentary
What people were saying around this time — external links, from the record's commentary rail.