Key Points
- Google introduced Gemini 4 Neon on November 18, aimed at complex coding, analysis and enterprise workflows.
- The model costs 32% less per task than Gemini 4 Carbon at its default high thinking level.
- Pricing and benchmark placement position Neon as Google's efficiency play against larger, slower frontier models.
The latest:
Google launched Gemini 4 Neon on November 18, a model built for software engineering, knowledge work and science workflows that the company says completes tasks 32% more cheaply than Gemini 4 Carbon at its default thinking level. Koray Kavukcuoglu, Senior Vice President at Google DeepMind and Chief AI Architect at Google, announced the release. It is available starting today.
Details:
- The benchmarks: Google said Neon scores 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, placing it third overall and ahead of comparable models, behind only larger and slower systems. On the Vals Index, the company said the model ranks second and sits on the Pareto frontier.
- The pricing: Google set Neon at $3 per million input tokens and $15 per million output tokens. Cached input tokens are priced at $0.15 per million, which the company described as a 95% discount. The model carries a 1 million-token context window.
- The cost claim: At the default high thinking level, Google said Neon combines a high Artificial Analysis Intelligence Index score with a 32% lower cost per task than Gemini 4 Carbon. At xhigh thinking, where it posts its strongest intelligence score, Google said it still costs at least 36% less per task than competing models at their highest setting.
- Where it ships: Google said Neon is rolling out today through the Gemini API in Google AI Studio, in Gemini for business via the API and the model dropdown, in Google Antigravity for paid Google AI Pro and Ultra subscribers, and in the Gemini app for Pro and Ultra subscribers.
- The security edition: A separate version, Gemini 4 Neon for cyber, is rolling out to trusted defenders in the coming days under what Google calls the Fairwind Program. Google did not name the participating organizations or set a specific date for that release.
- The lineup logic: Google positioned Neon as the middle tier of the Gemini 4 family, built to balance depth and efficiency. The company said Carbon handles the hardest long-horizon projects while Hydrogen covers fast everyday execution, leaving Neon for production-grade enterprise work.
- The target users: Google framed the model around complex coding, analytical reasoning and enterprise deployment rather than consumer chat, with availability weighted toward developers through the API and toward paying subscribers in Antigravity and the Gemini app.
- Not disclosed: Google did not publish the model’s parameter count, training data details or the comparison set behind the 36% figure, and did not name the competing models it measured against at their highest thinking setting.
Between the lines:
The headline claim Google is making is about cost per task rather than raw capability. Neon’s own benchmark placement is third overall, behind larger and slower models, so the company’s argument rests on the efficiency trade-off: comparable intelligence at a lower per-task price, with a tiered lineup that sends the hardest jobs to Carbon.
What’s next
Watch the rollout of Gemini 4 Neon for cyber to trusted defenders under the Fairwind Program in the coming days, and independent Artificial Analysis and Vals Index scoring once the model is live for developers.