Google announced a new AI model, Gemini 4 Neon ("Neon"), on November 18, 2026 (U.S. time) in a post on its official blog. The company says the model is built to complete complex software engineering, knowledge work, and science workflows quickly and end to end. It began rolling out the same day through the Gemini API for developers, paid plans of the Gemini app, and other products.

Neon Sits Between Carbon and Hydrogen

The announcement was published under the name of Koray Kavukcuoglu, SVP at Google DeepMind and Chief AI Architect at Google. Google's Gemini 4 lineup includes Gemini 4 Carbon, aimed at the hardest long-horizon work, and Gemini 4 Hydrogen, aimed at fast everyday execution. Neon slots in between, and Google positions it as a model that balances depth and efficiency.

The model offers four thinking levels. At the default "high" level, Google says its cost per task is 32% lower than Carbon's. The company also says Neon can act as an orchestrator in Google Antigravity, its development environment, directing multiple Hydrogen subagents in parallel.

API Pricing Starts at $3 per Million Input Tokens

API pricing is shown in the table below.

ItemPrice (per million tokens)
Input$3
Cached input$0.15
Output$15

Cached input is priced at a 95% discount to regular input. The context window, or the amount of input the model can handle at once, is 1 million tokens.

Strong Placements on Third-Party Benchmarks

Google cited results from several third-party evaluations. On the Artificial Analysis Intelligence Index, Neon scored 61 at its deepest "xhigh" thinking level, placing third overall. Google says that puts it just behind larger, slower models.

For coding, Google reports that Neon ranked first overall on the Artificial Analysis Coding Agent Index with 58.6 and scored 65.2% on Terminal-Bench 4.1. On business-oriented benchmarks, it placed second on the Vals Index at 70.9%, first on Zapier's AutomationBench at 56.3%, and first on Andon Labs' Vending-Bench 2 with $16,843. Google also says it ranked first on Legal Research Bench and Tax Agent Bench, though it did not give scores for those.

In science and math reasoning, Google reports 70.5% on Terminal-Bench Science, 84.0% on RiemannBench, and 71.3% on HLE-Diamond. All of the figures above were published by Google.

Paid Subscribers Get It in the Gemini App

Neon is rolling out from November 18 to the following:

  • The Gemini API in Google AI Studio (paid tier) and Android Studio
  • Gemini for business (via the API and the model drop-down)
  • Google Antigravity (paid Google AI Pro and Ultra subscribers)
  • The Gemini app (Google AI Pro and Ultra subscribers, via the model drop-down)

Google is also preparing Gemini 4 Neon for cyber, a security-focused version. The company says it will reach trusted defenders through its Fairwind Program over the next few days. The announcement does not mention regional availability or access for free users.

On safety, Google says Neon carries the same protections against severe misuse and prompt injection as Gemini 4 Carbon. On Gray Swan's indirect prompt injection evaluation, it recorded a 0.0% attack success rate at its highest thinking level, according to Google.

Our Take

Neon's main selling point appears to be near-top-tier performance at a lower cost. A 32% lower cost per task than Carbon should matter to developers and businesses running agents at scale. That said, the benchmarks were selected and published by Google itself, and how much difference the model makes in real work will depend on the use case. At $15 per million output tokens, it is also not cheap compared with lightweight models. Still, a new option between Carbon and Hydrogen is a welcome move for users who want to match models to their tasks.

Summary

Google has announced Gemini 4 Neon, a mid-tier model in its Gemini 4 generation, and started rolling it out through the API and paid plans of the Gemini app, among other products. It mainly matters to developers and businesses using the Gemini API, as well as Google AI Pro and Ultra subscribers. The security-focused Gemini 4 Neon for cyber is set to reach a limited set of organizations over the next few days.