Google Gemini 4 Neon Launches for AI Pro Users: Faster, Cost‑Effective Enterprise AI

Google’s newest Gemini 4 Neon model targets power users with enterprise‑grade reasoning, speed and a lower price point than its Carbon sibling. It’s now live in the Gemini app for AI Pro and Ultra subscribers.

Google Gemini 4 Neon Launches for AI Pro Users: Faster, Cost‑Effective Enterprise AI

Google has added a fresh model to its Gemini 4 family: Gemini 4 Neon. Aimed at developers, analysts and other heavy‑duty users, Neon promises the deep reasoning of the earlier Carbon model while delivering faster, lower‑cost execution. The model is already available in the Gemini app for AI Pro and AI Ultra subscribers, meaning anyone on those plans can start testing it today.

Key takeaways

  • Neon balances the deep reasoning of Carbon with the speed and lower cost of Hydrogen.
  • It is accessible to AI Pro users, not just the higher‑tier AI Ultra accounts.
  • Pricing is $3 per million input tokens and $15 per million output tokens, with discounts for cached inputs.
  • Neon offers a 1 million‑token context window and four “thinking levels” (Low, Medium, High, Extra High).
  • Benchmarks place Neon in the top three across multiple metrics, including legal research, finance and code generation.

What Neon is and why it matters

Gemini 4 Neon sits between Hydrogen and Carbon in Google’s “elemental” model lineup. Carbon is known for deep, long‑horizon reasoning, while Hydrogen excels at rapid, everyday tasks. Neon is positioned as a middle ground: it can handle complex, high‑frequency workflows—such as large‑scale code generation or intricate data analysis—while running faster and at a lower price point than Carbon.

This balance matters for businesses and power users who need both depth and speed. The model’s design targets enterprise workloads, scientific research, and software development, offering the flexibility to stay efficient on prolonged tasks without sacrificing output quality.

Performance and benchmark highlights

According to Google’s own testing, Neon lands in the top three on several industry benchmarks. Highlights include:

  • #1 on the Legal Research Bench and Tax Agent Bench.
  • #2 on the Agent Arena Index for steerability (ability to adapt to mid‑task corrections).
  • #1 on Zapier’s AutomationBench and Andon Labs’ Vending‑Bench 2.
  • Strong scores on finance, accounting, spreadsheet modeling and constraint‑following tasks.
  • Chartography reading accuracy of 73.5% for dense professional charts.

These results show Neon’s ability to outperform many competing models, including those from Anthropic and OpenAI, especially in tasks that require both reasoning depth and rapid iteration.

How Neon is priced

Token typePrice (per million tokens)
Input tokens$3
Output tokens$15

Google also offers discounts for cached input tokens, making repeat queries even cheaper. By comparison, Hydrogen costs half as much per token, while Carbon and Argon are priced at $2 input / $10 output until November 30, after which they rise to $4/$20. Neon’s pricing is therefore competitive for high‑frequency, enterprise‑scale usage.

Availability and how to start using Neon

Neon is live in the Gemini app for users on the AI Pro plan and for AI Ultra subscribers. To enable the model, open the Gemini app, go to the model selector, and choose “Neon.” The model replaces Argon in the picker, and the previous 3.1 Pro version is being retired. Users on paid plans also see the 3.5 Flash‑Lite model set to “Auto.”

Because Neon works with the same 1 million‑token context window as Hydrogen, existing prompts and workflows that already fit within that limit can be switched to Neon without major changes. The four thinking levels let you trade speed for depth as needed:

  • Low – fastest, suitable for simple queries.
  • Medium – balanced speed and reasoning.
  • High – deeper analysis for more complex tasks.
  • Extra High – maximum reasoning power for the toughest problems.

What this means for you

If you already pay for AI Pro, you can start testing Neon immediately. Consider the following steps:

  1. Open the Gemini app and select Neon from the model dropdown.
  2. Choose the appropriate thinking level based on your workload (e.g., High for code reviews, Medium for data summaries).
  3. Monitor token usage in your account dashboard to gauge cost impact.
  4. Experiment with cached inputs for repetitive prompts to maximize the discount.

For teams, Neon’s lower price per token can reduce overall AI spend while still delivering Carbon‑level quality on mission‑critical tasks.

What happens next

Google has indicated that Neon is part of an ongoing “elemental” expansion, so additional models or refinements are likely. The company also plans to keep the pricing promotional until the end of November, after which Neon will become 25 % cheaper relative to the post‑promo rates for Carbon and Argon. Users should watch for any pricing updates and for possible new features in the Gemini app, such as expanded context windows or additional thinking levels.

Overall, Gemini 4 Neon gives AI Pro users a powerful, cost‑effective tool for enterprise‑grade workloads, positioning Google as a strong contender in the high‑performance AI market.

Frequently asked questions

How can I access Gemini 4 Neon?

Neon is available in the Gemini app for AI Pro and AI Ultra subscribers; just select it from the model picker.

What is Neon’s token pricing?

Neon costs $3 per million input tokens and $15 per million output tokens, with discounts for cached inputs.

What are the thinking levels in Neon?

Neon offers Low, Medium, High and Extra High settings, letting you balance speed and reasoning depth.

How does Neon compare to Carbon?

Neon matches Carbon’s deep reasoning but runs faster and at a lower cost, making it suitable for high‑frequency tasks.

Is Neon suitable for enterprise workflows?

Yes, Google markets Neon for complex coding, analytical reasoning and production‑grade enterprise use cases.

Sources

This story was prepared with the help of AI tools, based on the reporting listed above. Details of developing stories may change.