Who can use it
Google announced Gemini 4 Neon on 18 November for coding, analysis and knowledge work. The rollout includes the paid Gemini API, Gemini for business and eligible Google AI Pro and Ultra subscribers in the Gemini app and Antigravity. The company describes Neon as balancing demanding work with faster execution.
Standard API rates are $3 per million input tokens and $15 per million output tokens, with cached input at $0.15. Tokens are the units a model reads and generates. These rates do not tell you the price of a finished task: a longer answer or a more involved workflow can consume more tokens.
A separate route for cyber work
Google says Neon for cybersecurity is reaching trusted defenders through its Fairwind programme over the following days. That is a separate access statement, not a promise that every subscriber receives the same cyber capabilities immediately. The launch also offers four levels of reasoning effort, allowing users to trade additional processing against task demands.
What the published evidence does not settle
The model card says Neon is adapted from Gemini 4 Carbon and accepts text, images, audio and video within a combined context of up to one million tokens. Context is the material a model can work with in a session; capacity alone does not guarantee it will interpret every detail correctly.
Google explicitly lists hallucinations among the limitations and gives a March 2026 knowledge cutoff. Its card describes some automated safety tests, not human evaluation or red-teaming results, and still lists a full Neon Safety and Security Report as forthcoming. Several frontier-risk determinations are carried forward from Gemini 4 Argon under Google's assessment framework.
For users, the practical distinction is between a newly available tool and a fully established account of its reliability. Google's claims about speed, cost and benchmark performance are not independent Model Current tests. Important outputs still need checking against their underlying evidence.
Sources
- Google: Introducing Gemini 4 Neon18 November launch, access channels, standard token prices and the separate Fairwind rollout.
- Google DeepMind: Gemini 4 Neon model cardModel inputs, limits, training cutoff, evaluation scope and forthcoming safety report.



