Connect with us

Google Launches Gemini 4 Neon: A New Mid-Tier AI Model

Google Launches Gemini 4 Neon

Credit: Shutterstock

Meet Gemini 4 Neon, Google’s new AI model that claims to beat Claude Opus 5.5 and GPT-6.2 Sol.

Google is not slowing down in the AI race. The company officially launched Gemini 4 Neon on November 18, 2026, adding a fourth model to its Gemini 4 family. The release came just over a week after Gemini 4 Carbon and Hydrogen arrived.

Before today, Google’s Gemini 4 family consisted of Hydrogen for everyday help, Argon for code and analysis, and Carbon for the most complex work. Gemini 4 Neon fills the gap between Hydrogen and Carbon. Google says the mid-tier model brings frontier-level intelligence to complex coding, knowledge work, and science without the long wait. In short, it aims to deliver Carbon-level performance at Flash speeds and a lower price.

The model is now available in the Gemini app for Google AI Pro and Ultra subscribers, replacing Gemini 4 Argon and the older Gemini 3.1 Pro. The updated model picker now gives subscribers four options:

  1. Auto
  2. Hydrogen
  3. Neon
  4. Carbon

Neon is also available through the paid Gemini API in Google AI Studio and Android Studio. It is listed in the Vertex AI documentation for enterprise customers in the global, US, and EU regions.

API pricing is set at $3 per million input tokens and $15 per million output tokens, with cached input at $0.15. The context window is one million tokens, and the model has four thinking levels, with ‘High’ as the default. Neon is priced twice as high as Hydrogen. However, at its default ‘High Thinking’ level, Google says Neon delivers tasks at 32% lower cost per task than Carbon, and 50 to 80% less than other leading frontier models. Those figures are Google’s own.

Google’s Logan Kilpatrick shared the news on X shortly after launch:

On benchmarks, Google’s own comparison table puts Neon ahead of Claude Opus 5.5 and GPT-6.2 Sol on the Artificial Analysis Intelligence Index, scoring 61 against 59 and 57 respectively. Google also says Neon ranks first on the Legal Research Bench and Tax Agent Bench. These numbers, however, come from Google, and independent evaluations have not yet been published.

A separate version of the model, Gemini 4 Neon for cyber, is rolling out to trusted security defenders through Google’s Fairwind Program over the coming days.

The model card, published November 18, lists a knowledge cutoff of March 2026, with some domains less current than others.

Google has been working to close the gap with Anthropic and OpenAI, but months of delays to its AI model have caused the company to fall behind both rivals. The rapid rollout of the Gemini 4 lineup, with four models all named after chemical elements, Argon, Carbon, Hydrogen, and now Neon, suggests Google is pushing hard to make up for lost time. If Neon’s numbers hold up under independent testing, it could mark a real turning point for the company in this competition.