The short answer
- Google announced Gemini 4 Neon on November 18, 2026, and said the model is rolling out in the Gemini API, Gemini for business, Google Antigravity and the Gemini app.
- Google prices Gemini 4 Neon at $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens, on the paid tier only, with a 1 million-token context window.
- Through November 30, 2026, Neon costs more per token than Gemini 4 Carbon and Gemini 4 Argon at their $2 input and $10 output rates. From December 1, 2026, those models rise to $4 and $20, 25% above Neon’s flat rate.
- Google says Neon scores 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, and 39.0% on SWE-sweep in the Google Antigravity harness, compared with 12.8% on the mini-SWE harness.
- GitHub said on November 18, 2026, that Gemini 4 Neon is generally available in GitHub Copilot for Pro+, Max, Business and Enterprise users, with a gradual rollout.
Google announced Gemini 4 Neon on November 18, 2026, and said the model is rolling out in the Gemini API in Google AI Studio, Gemini for business, Google Antigravity and the Gemini app. In a post on The Keyword, Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect at Google, wrote: “The model is priced at $3 per million input tokens ($0.15 per million cached input tokens, a 95% discount) and $15 per million output tokens, with a 1 million-token context window.” Paid Google AI Pro and Ultra subscribers can select it in the Gemini app and try it in Google Antigravity. Jay Peters of The Verge also reported the announcement on November 18, 2026.
#What changed?
Google says: “Neon is built for complex coding, analytical reasoning, and production-grade enterprise workflows. While Carbon tackles your hardest, long-horizon projects and Hydrogen handles fast everyday execution, Neon balances depth and efficiency.” Google’s model card adds: “Gemini 4 Neon is adapted from Gemini 4 Carbon.”
Google’s Gemini API release notes list gemini-4-neon as generally available on November 18, 2026. They say it “supports four thinking levels (low, medium, default high, and xhigh), a unified 1M-token context window, and long decode up to 1M tokens.” Google’s models page lists a maximum output of 64K or 1M tokens, with text, image, audio, video and PDF modalities. Google Cloud’s model page lists a March 2026 knowledge cutoff. Google also says Tasks in Gemini can run on Neon, “working through complex, multi-step jobs more quickly and efficiently than with larger models.”
Google says Neon “ranks ahead of comparable models and #3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, landing just behind larger, slower models.” It also says: “Neon ranks #2 and lands on the Pareto frontier.” Google’s comparison table lists that Vals Index result at 70.9%, against Claude Opus 5.5 and GPT-6.2 Sol. Google’s evals methodology says: “All the results for non-Gemini models are sourced from providers’ self-reported numbers unless otherwise mentioned below.”
Google says Neon ranks first on the Artificial Analysis Coding Agent Index at 58.6, with 65.2% on Terminal-Bench 4.1 and 46.2% at its default high setting on FrontierCode 2.0. Google’s own table shows Claude Opus 5.5 ahead on both tests, at 65.8% and 47.3%. On Chartography, Google lists Neon at 73.5% and GPT-6.2 Sol ahead at 75.1%.
On SWE-sweep, Google says: “it already leads on the mini-SWE harness (12.8%) and more than triples to a state-of-the-art 39.0% in Google Antigravity showing what the Antigravity harness unlocks.” Google’s methodology says that 39.0% figure uses “the Antigravity-CLI version 1.5.9 harness through official integration for Gemini 4 Neon, Claude Code for Opus 5.5, and Codex for Sol 6.2.” In that harness, Google lists Claude Opus 5.5 at 10.1% and GPT-6.2 Sol at 8.3%.
Google also says: “At its default high thinking level, it pairs a high Artificial Analysis Intelligence Index score with a 32% lower cost per task than Carbon.” Thai Tran, AI Product Lead at Glean, is quoted in Google’s post: “In our early evaluations, it matched or exceeded Gemini 4 Carbon on overall quality at 3x the speed and 60% of the cost.” That is a customer evaluation quoted by Google.
#How much does it cost?
Google’s paid-tier price is $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens, flat across the 1 million-token context window. As of November 19, 2026, the Gemini API pricing page lists the same rates and says output includes thinking tokens. The published Neon rate is one flat price. The page says: “Gemini 4 Neon is not available on the Free tier.” Caching is implicit only, and explicit caching is not supported. Google Cloud’s documentation lists a 4,096-token minimum for implicit caching.
On the same pricing page, Carbon and Argon are $2.00 input and $10.00 output per million tokens through November 30, 2026, then $4.00 and $20.00 starting December 1, 2026. Through November 30, Neon costs more per token than those models. From December 1, Neon’s $3 and $15 rates are 25% below the scheduled Carbon and Argon rates. Gemini 4 Hydrogen is $1.50 input and $7.50 output, and Google’s release notes also place it on the free tier. GitHub’s Copilot changelog says the model “is billed at provider list pricing under usage-based billing.”
#Who can use it and when?
On Google Cloud, Google says: “Gemini 4 Neon is available to verified enterprise customers in select regions.”
Surface | What the sources say |
|---|---|
Gemini API, Google AI Studio and Android Studio | Paid-tier API, per Google. Pro and Ultra subscribers can try Neon in AI Studio. |
Google Antigravity | Paid Google AI Pro and Ultra subscribers, per Google. |
Gemini app | Pro or Ultra subscribers can select Neon from the model drop-down. |
Gemini for business | Via the API or the model drop-down, and in Google Antigravity, per Google. |
Google Cloud | Verified enterprise customers in select regions (global, us, eu). |
Tasks in Gemini | Google says Tasks can run on Neon. |
GitHub Copilot | Generally available to Pro+, Max, Business and Enterprise users, GitHub says, with a gradual rollout. |
Gemini 4 Neon for cyber | Eligible Fairwind Program members, after the November 18, 2026 announcement, per Google. |
GitHub says Copilot Enterprise and Business administrators can manage access through the model policy in Copilot settings.
#How to get it
- In the Gemini API paid tier, use the model ID
gemini-4-neonin Google AI Studio or Android Studio. - Google AI Pro or Ultra subscribers can select Neon from the Gemini app model drop-down, and can try it in Google Antigravity and in AI Studio.
- For Gemini for business, Google points teams to the API or the model drop-down.
- On GitHub Copilot, eligible users can select the model in the model picker as the gradual rollout reaches them.
- To request Fairwind Program access, Google says: “contact your Cloud account team.”
#How does it compare?
Gemini 4 Carbon first went to cyber defenders, and Gemini 4 Hydrogen launched for fast everyday execution. Google still assigns Carbon the hardest long-horizon projects and Hydrogen fast everyday execution, with Neon as the balance of depth and efficiency.
Our analysis: Neon’s headline token price is higher than Carbon’s and Argon’s through November 30, 2026, and 25% lower from December 1, 2026. Google’s “32% lower cost per task than Carbon” claim is a Google figure at the default high thinking level, not the same as the headline token rate. The benchmark numbers are also Google’s claims, and Google’s own table shows Claude Opus 5.5 or GPT-6.2 Sol ahead on three tests.
#What is still unclear?
When will eligible Fairwind Program members receive Gemini 4 Neon for cyber? When will the GitHub Copilot rollout reach all Pro+, Max, Business and Enterprise users? What will Google Cloud charge for Neon on the Gemini Enterprise Agent Platform? How will Neon appear in the Gemini app model picker next to other Gemini 4 models?
Frequently asked questions
What is Gemini 4 Neon?
Gemini 4 Neon is a Gemini 4 model Google announced on November 18, 2026. Google says it is built for complex coding, analytical reasoning and production-grade enterprise workflows, and that it balances depth and efficiency between Carbon and Hydrogen. Google’s model card says Neon is adapted from Gemini 4 Carbon.
How much does Gemini 4 Neon cost?
Google prices the model at $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens, with a 1 million-token context window. The rate is flat, and output includes thinking tokens. Google’s pricing page says Neon is not available on the free tier. Through November 30, 2026, Carbon and Argon cost $2 and $10; from December 1, 2026, they rise to $4 and $20.
Who can use Gemini 4 Neon?
Google says the model is rolling out in the Gemini API paid tier in Google AI Studio and Android Studio, in Gemini for business, and for paid Google AI Pro and Ultra subscribers in Google Antigravity and the Gemini app. On Google Cloud, it is available to verified enterprise customers in select regions. GitHub said it is generally available in Copilot for Pro+, Max, Business and Enterprise users, with a gradual rollout. Google said eligible Fairwind Program members would begin receiving the cyber version after the November 18, 2026 announcement.
How does Gemini 4 Neon compare with Carbon and Hydrogen?
Google says Carbon tackles the hardest long-horizon projects, Hydrogen handles fast everyday execution, and Neon balances depth and efficiency. Hydrogen is priced at $1.50 input and $7.50 output per million tokens. Neon’s $3 and $15 rates are higher than Carbon’s and Argon’s $2 and $10 through November 30, 2026, and 25% lower than their $4 and $20 rates from December 1, 2026.
How do developers start using Gemini 4 Neon?
Google’s release notes list the model ID as gemini-4-neon and say it is available on the paid tier in the Gemini API. Google tells developers to build with that paid tier in Google AI Studio and Android Studio. Paid Google AI Pro or Ultra subscribers can try Neon in Google Antigravity and in AI Studio. GitHub Copilot users on eligible plans can select it in the model picker as the gradual rollout reaches them.
Sources
Every factual claim in this article traces back to one of these.
One email a week on AI agents and bots.
Launches, pricing and rollout changes and the guides worth reading. At most once a week, and you can leave with one click.




