Gemini 4 Neon Is Here: Google's New Coding Model

Duotone photo of a code editor on a glowing monitor in indigo and lime green, with the words "NEON CODES FASTER."

Google has added a fourth model to its Gemini 4 line-up in a month. On 18 November 2026, the company launched Gemini 4 Neon, built for complex software engineering, knowledge work and science workflows.

Neon is the middle option: production-ready code and long-running agent workflows without paying for the heaviest model. Here is what it does, what it costs, and where you can try it.

Where Neon sits in the Gemini 4 family

Google has now released four Gemini 4 models in about five weeks: Argon on 13 October, Carbon on 10 November, Hydrogen on 12 November, and Neon on 18 November. Each is aimed at a different job, trading power against speed.

Carbon is the heaviest model, built for long-running projects that need the most capability. Hydrogen is the fast everyday model, for quick tasks where speed matters more than depth. Neon sits between them: stronger than Hydrogen on hard work, cheaper and lighter than Carbon for jobs that still need serious reasoning.

That middle position is the point. Much real software work is neither a five-minute question nor a week-long project — a feature that takes an afternoon, or a refactor across dozens of files. Neon is built for that middle ground.

What Neon costs — and the December price twist

Neon costs $3 per million input tokens and $15 per million output tokens through the Gemini API. Cached input costs $0.15 per million tokens, a 95% discount, and the output price includes thinking tokens, so the reasoning before an answer is billed as output.

The model also comes with a 1,048,576-token unified context window — roughly a million tokens in one go, enough to load a big codebase and long documents into a single session.

From 1 December, the picture changes. Carbon and Argon cost $2 input and $10 output per million tokens through 30 November 2026, then rise to $4 and $20 from 1 December. From then, Neon at $3 and $15 is 25% cheaper per token than Carbon.

Google also says Neon beats Carbon on core accuracy benchmarks while cutting per-task costs by 17% overall, and by 32% in high-thinking mode, even though its token rates are 1.5 times Carbon's introductory rates.

What Google's benchmarks say

All benchmark figures below are Google's own claims, reported at launch. Vendor benchmarks help compare models on the same tests, but independent testing often tells a different story, so treat these numbers as a starting point, not a verdict.

According to Google, Neon ranks third overall with a score of 61 on the Artificial Analysis Intelligence Index at xhigh thinking. On coding, it takes first place on the Artificial Analysis Coding Agent Index with 58.6 — the headline result for a model aimed at developers.

Google reports 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0 at high thinking, second place on steerability on the Agent Arena Index, and first place on most Surge AI finance and accounting benchmarks.

Google reports 76% on the hard tier of AutoBaxBench, which tests secure coding, and calls that state-of-the-art. It also says Neon has the lowest fabrication rate yet in its internal Antigravity testing — it invents fewer false details than earlier models.

Where you can use Neon

Neon has been rolling out since 18 November in the Gemini API in Google AI Studio, Gemini for business, Antigravity and the Gemini app for Pro and Ultra subscribers. The model ID is gemini-4-neon. It is also in GitHub Copilot for Pro+, Max, Business and Enterprise plans, putting it inside the editor many developers already use.

In Antigravity, Neon can act as an Orchestrator, directing parallel Hydrogen subagents. A defence-tuned variant is available to vetted defenders through the Fairwind Program, aimed at security work rather than general coding.

Key takeaways

  • Google launched Gemini 4 Neon on 18 November 2026, the fourth Gemini 4 model in a month, sitting between Carbon and Hydrogen.
  • It costs $3 per million input tokens ($0.15 cached) and $15 per million output tokens, with a 1,048,576-token context window.
  • Google claims first place on the Artificial Analysis Coding Agent Index (58.6), 65.2% on Terminal-Bench 4.1, and 76% on the AutoBaxBench hard tier for secure code.
  • Google says Neon beats Carbon on core accuracy benchmarks while cutting per-task costs by 17% overall and 32% in high-thinking mode; from 1 December, Neon is 25% cheaper per token than Carbon.
  • Neon is rolling out in the Gemini API, Gemini for business, Antigravity and the Gemini app for Pro and Ultra subscribers, and in GitHub Copilot.

Conclusion

Neon fills a gap Google clearly thinks matters: strong enough for serious coding and agent work, but not so heavy that every task costs top price. The benchmark claims are strong, especially on coding — but the numbers are Google's own, and the real test will be independent results.

If you already use the Gemini API, GitHub Copilot or a Google AI Pro or Ultra plan, try Neon on a real project. Compare it with what you use now, watch the total cost per finished job rather than the price per token, and decide from there.

FAQ

What is Gemini 4 Neon?

Gemini 4 Neon is Google's coding-focused Gemini 4 model, launched on 18 November 2026. It sits between Carbon, the heaviest model, and Hydrogen, the fast everyday model.

How much does Gemini 4 Neon cost?

Neon costs $3 per million input tokens and $15 per million output tokens. Cached input costs $0.15 per million tokens, and the output price includes thinking tokens. From 1 December 2026, when Carbon's price rises to $4 input and $20 output, Neon will be 25% cheaper per token than Carbon.

Where can I use Gemini 4 Neon?

Neon is rolling out in the Gemini API in Google AI Studio, Gemini for business, Antigravity and the Gemini app for Pro and Ultra subscribers, and in GitHub Copilot for Pro+, Max, Business and Enterprise plans. The model ID is gemini-4-neon.

Sources

Researched with AI assistance and reviewed for accuracy.

Follow on WhatsApp

Follow Daily Tech & AI on WhatsApp for one tech and AI update every day: Daily Tech & AI WhatsApp channel.

Comments