Google released Gemini 4 Neon on November 18, 2026, and it is now the highest-scoring Gemini model. Neon scores 61 on the Artificial Analysis Intelligence Index, one point above Gemini 4 Carbon. Neon costs $3 per million input tokens and $15 per million output tokens on the paid Gemini API. In the Gemini app, it is the model Google AI Pro subscribers now get in place of Gemini 4 Argon.
That last change is the part most launch coverage skips. Google's own help page now lists only three Gemini 4 models in the app, Hydrogen, Neon and Carbon, and Argon is gone from every plan. This guide covers what Gemini 4 Neon is and who gets it on each plan. It shows what Neon costs against Carbon before and after the November 30 price change, and where its benchmarks hold up. It also shows the three rows of Google's own table where Claude Opus 5.5 and GPT-6.2 Sol still win.
The Key Takeaways
- Released November 18, 2026: Gemini 4 Neon is adapted from Gemini 4 Carbon and sits between Hydrogen and Carbon in the Gemini 4 family.
- Who gets it: Google AI Pro ($19.99 a month) and Ultra in the Gemini app. Free and Plus users get Hydrogen. Argon is no longer in the app on any plan.
- Price: $3 / $15 per million tokens, with no free API tier. That is 50% more than Carbon's introductory $2 / $10 until November 30, and 25% less than Carbon's $4 / $20 from December 1.
- Benchmarks: 61 on the Artificial Analysis Intelligence Index v5.0.1 at xhigh, against 60 for Carbon at high and 54 for Hydrogen at xhigh.
- Where it loses: in Google's own table, Claude Opus 5.5 beats Neon on Terminal-Bench 4.1 (65.8% vs 65.2%) and FrontierCode 2.0, and GPT-6.2 Sol beats it on Chartography.
What Is Gemini 4 Neon?
Gemini 4 Neon is Google's mid-tier Gemini 4 model, built for complex coding, analysis and long agent workflows that need to finish quickly. Google's model card says Neon is adapted from Gemini 4 Carbon, the same way Hydrogen was. Google describes the split in one line: Carbon tackles the hardest long-horizon projects, Hydrogen handles fast everyday execution, and Neon balances depth and efficiency.
Neon is the fourth Gemini 4 model released since October 13. Gemini 4 Argon went public first, Gemini 4 Carbon followed on November 10, and Gemini 4 Hydrogen on November 12. The model ID is gemini-4-neon, and it shipped as a stable, generally available model on day one, not a preview.
Gemini 4 Neon Specs
Gemini 4 Neon's core specs match the rest of the Gemini 4 family. Gemini 4 Neon has a 1,048,576-token context window shared by input, thinking and output. It returns up to 65,536 tokens per synchronous response, and up to the full 1M with Google's long-decode mode. It reads text, images, video, audio and PDFs, and it writes text only: no image or audio generation. The model card gives a knowledge cutoff of March 2026.
Google announced the model from its main AI account with a demo of a 12-hour agent run in Antigravity:
Meet Gemini 4 Neon — our newest model that brings frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows.
— Google AI (@GoogleAI) November 18, 2026
Neon brings frontier-level performance to your everyday work and is built to handle complex coding and problem-solving without the wait.
How, you might ask? ⬇️
— Smarter coding: Multi-file debugging and refactoring with targeted, surgical edits
— Grounded answers: Excels at multi-step instruction following and builds on our ongoing progress to curb fabrication, recording our lowest fabrication rate yet in internal Google Antigravity testing
— 1M-token context window: Synthesize massive projects, tables, and charts at once
— Adjustable thinking: Balance and adjust speed and depth on the fly with 4 thinking levels
Watch Gemini 4 Neon run a 12-hour /swarm in @antigravity to reconstruct Alan Turing’s 1936 study from archival photos — complete with a running Universal Turing Machine, the blueprint for modern computing.
Who Can Use Gemini 4 Neon?
Gemini 4 Neon is available to Google AI Pro and Ultra subscribers in the Gemini app. Developers get it on the paid tier of the Gemini API, and Pro and Ultra subscribers also get it in Google AI Studio and Google Antigravity. Free Gemini users and Google AI Plus subscribers do not get Neon.
Gemini 4 Neon in the Gemini App, Plan by Plan
Google's help page About Gemini models, updated for the Neon launch, lists the models each plan gets in chats and in tasks:
| Gemini plan | US price | Models in chats | Models in tasks |
|---|---|---|---|
| No plan | Free | Auto only | Hydrogen |
| Google AI Plus | $4.99 / month | Auto, Hydrogen | Hydrogen |
| Google AI Pro | $19.99 / month | Auto, Hydrogen, Neon | Hydrogen, Neon |
| Google AI Ultra | from $99.99 / month | Auto, Hydrogen, Neon, Carbon | Hydrogen, Neon, Carbon |
Free users can only chat through Auto, and Plus is the cheapest plan that lets you pick Hydrogen yourself. Google's help page says Auto picks between Hydrogen and "a fast, lightweight model" for each prompt, so Auto never reaches Neon. To use Gemini 4 Neon, a Pro or Ultra subscriber has to select it in the model menu.
What Happened to Gemini 4 Argon?
Gemini 4 Argon is no longer a model you can pick in the Gemini app. The help page lists Hydrogen, Neon and Carbon and nothing else, and Google AI Pro's plan card now names Neon and Hydrogen. 9to5Google reported the same change on launch day: "Neon also replaces Argon."
Argon has not disappeared from Google's developer platform. The Gemini API changelog has no deprecation notice for gemini-4-argon, and the API pricing page still lists it at the same $2 / $10 introductory rate as Carbon. Apps and tools that call Argon through the API keep working for now. If you used Argon in the Gemini app as a Pro subscriber, Neon is the model that took its slot.
Gemini 4 Neon Thinking Levels: App vs API
The Gemini app offers Gemini 4 Neon at three thinking levels, Low, Medium and High. Google's help page says a manual model pick defaults to Low unless you change it. The API adds a fourth level, xhigh, and defaults to High. The top benchmark score Google advertises, 61 on Artificial Analysis, was measured at xhigh, a setting app users cannot select.
For Developers: API, AI Studio and Antigravity
Developers can call Gemini 4 Neon through the Gemini API paid tier, in both generateContent and the Interactions API. Google says all Pro and Ultra subscribers can try it in Google AI Studio. Paid Pro and Ultra subscribers also get it in Google Antigravity, where Neon can run alone or direct parallel Hydrogen subagents. Businesses get Neon in Gemini for business, through the API or the model drop-down. A cybersecurity version, Gemini 4 Neon for cyber, goes to vetted defenders in Google's Fairwind Program over the next few days.
The Gemini app's own account confirmed the consumer rollout on launch day:
Starting today, Gemini 4 Neon is available to Google AI Pro and Ultra subscribers on mobile, desktop, and web.
— Google Gemini (@GeminiApp) November 18, 2026
This new model excels at work that requires frontier-level reasoning and fast responses. 🧵
Gemini 4 Neon Pricing
Gemini 4 Neon costs $3 per million input tokens and $15 per million output tokens on the Gemini API, including thinking tokens. Cached input costs $0.15 per million, a 95% discount, through implicit caching only. Neon has no free tier: every token is billed. Batch and Flex requests cost half, $1.50 / $7.50.
| Model | Through Nov 30, 2026 | From Dec 1, 2026 | Free API tier |
|---|---|---|---|
| Gemini 4 Hydrogen | $1.50 / $7.50 | $1.50 / $7.50 | Yes |
| Gemini 4 Neon | $3 / $15 | $3 / $15 | No |
| Gemini 4 Carbon | $2 / $10 | $4 / $20 | No |
| Gemini 4 Argon | $2 / $10 | $4 / $20 | No |
The table explains an odd launch-week fact. Until November 30, Neon costs 50% more per token than Carbon, Google's strongest model, because Carbon is still on its introductory rate. From December 1, Carbon doubles and Neon becomes 25% cheaper. A team choosing between the two on price alone should run Carbon this month and switch the comparison on December 1. Our Gemini pricing guide covers the consumer plans and Vertex AI rates.
Per-token price is not the whole bill. Artificial Analysis found that Neon at xhigh produced 283 million output tokens across its Intelligence Index, which it calls "very verbose". Its cost per index task was $6.44 for Neon (xhigh) against $4.90 for Carbon (high) at today's prices. At the default high setting the gap reverses: Artificial Analysis lists Neon (high) at 59.6 on the index for $3.74 per task, 24% below Carbon (high) at today's prices. Google's launch post claims a 32% saving at high.
Gemini 4 Neon Benchmarks
Gemini 4 Neon is Google's best-scoring model on independent tests, but by a small margin, and it does not win every row of Google's own comparison table.
Artificial Analysis Scores
On the Artificial Analysis Intelligence Index (v5.0.1), Gemini 4 Neon scores 61 at xhigh. Gemini 4 Carbon scores 60 at high, and Gemini 4 Hydrogen scores 54 at xhigh. Google's blog says Neon ranks third overall, behind larger, slower models.
Speed cuts both ways. Artificial Analysis measures Neon's output speed at 110 tokens per second, against 57 for Carbon. Yet it also rates Neon "notably slow per task", at 21.5 minutes per index task against a 12.4-minute median, because Neon at xhigh writes so many tokens. Fast tokens, long answers: plan for both.
Where Neon Wins and Loses in Google's Own Table
Google's model card compares Gemini 4 Neon with Claude Opus 5.5 and GPT-6.2 Sol. Neon leads most rows, but not all of them:
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| AA Intelligence Index v5.0.1 | 61 | 59 | 57 |
| Vals Index v2.2 | 70.9% | 70.1% | 66.1% |
| Terminal-Bench 4.1 | 65.2% | 65.8% | 61.5% |
| FrontierCode 2.0 | 46.2% | 47.3% | 43.8% |
| CursorBench 4.1 | 65.4% | 58.1% | not listed |
| Agents' Last Exam v1.1 | 75.8% | 71.6% | 73.0% |
| HLE-Diamond | 71.3% | 67.9% | 64.0% |
| Chartography | 73.5% | 66.3% | 75.1% |
| GraphWalks, 256K to 1M | 95.2% | 66.8% | 80.1% |
Two patterns stand out. On agentic coding, Neon and Opus 5.5 are close to a tie: Opus wins Terminal-Bench 4.1 and FrontierCode 2.0 by about a point, and Neon wins CursorBench 4.1 by seven. On long context, Neon is far ahead, at 95.2% on GraphWalks across 256K to 1M tokens. The table compares Neon with Opus 5.5 and Sol, not with Claude Fable 5.2 or GPT-6.2 Astra, both of which score higher on the Artificial Analysis index. For the full field, see our Gemini 4 tracker.
Gemini 4 Neon vs Hydrogen vs Carbon: Which One to Use
Use Gemini 4 Neon for complex work you run often: multi-file coding, long documents, spreadsheets and research reports. Use Carbon for the hardest single problems, and Hydrogen for everyday speed. Google's developer guide gives the same split.
| Hydrogen | Neon | Carbon | |
|---|---|---|---|
| Best for | Everyday chat, fast coding | Complex, frequent work | Hardest long-horizon problems |
| AA Index | 54 (xhigh) | 61 (xhigh) | 60 (high) |
| API price | $1.50 / $7.50 | $3 / $15 | $2 / $10, then $4 / $20 |
| Cheapest app plan | Free (via Auto), Plus to pick it | Google AI Pro | Google AI Ultra |
For a Pro subscriber, the choice is simple. Neon is the strongest model on Google AI Pro, so pick it for anything that needs reasoning and leave Auto for quick questions. Want Gemini's answer next to Claude's or ChatGPT's on the same prompt, without paying for three subscriptions? Fello AI keeps Gemini, Claude and GPT in one Mac app, so switching models is a click, not a new account.
What the Gemini 4 Neon Model Card Adds
The Gemini 4 Neon model card adds three facts the launch post does not. Neon is adapted from Carbon, so its training data, hardware and acceptable-use rules all point back to the Carbon card. Its knowledge cutoff is March 2026. And the full Safety and Security Report is not out yet: Google says it will be available "in the coming days".
Google's launch post does publish some safety numbers. It says Neon's cyber and biosecurity safeguards wrongly declined just 1.98% of legitimate tasks on Terminal-Bench 4.0, and that Neon passed 99.2% of Google's agent honesty evaluation. Neon took shortcuts on 2.45% of tasks in Google's coding cheating test, the lowest rate of any Gemini 4 model. These are Google's internal tests, so treat them as Google's claims until the full report lands.
The Verdict on Gemini 4 Neon
Gemini 4 Neon is the model most Gemini subscribers will actually use. It is the strongest model on Google AI Pro, it took Argon's slot in the app, and it edges Carbon on the Artificial Analysis index. The catch is timing: until November 30, Carbon is cheaper on the API, and app users cannot reach the xhigh setting behind Neon's best score. For Pro subscribers, switch the picker to Neon today. For developers, test it now and decide on December 1, when Carbon's price doubles and Neon's does not.
Frequently Asked Questions
Is Gemini 4 Neon free?
No. Gemini 4 Neon is not on the Gemini API free tier, and in the Gemini app it needs Google AI Pro ($19.99 a month) or Google AI Ultra. Free and Google AI Plus users get Gemini 4 Hydrogen instead.
When was Gemini 4 Neon released?
Google released Gemini 4 Neon on November 18, 2026, as a generally available model on the Gemini API and in the Gemini app for Pro and Ultra subscribers on the same day.
Is Gemini 4 Neon better than Gemini 4 Carbon?
Slightly, on one index. Gemini 4 Neon scores 61 on the Artificial Analysis Intelligence Index at xhigh against 60 for Carbon at high. Google still positions Carbon as its model for the hardest long-horizon problems, and Carbon is cheaper on the API until November 30.
Is Gemini 4 Argon gone?
Gemini 4 Argon is gone from the Gemini app, where Google's help page now lists only Hydrogen, Neon and Carbon. It is still listed on the Gemini API at $2 / $10 until November 30 and $4 / $20 from December 1, with no deprecation notice so far.
How much does the Gemini 4 Neon API cost?
The Gemini 4 Neon API costs $3 per million input tokens and $15 per million output tokens, with cached input at $0.15. Batch and Flex requests cost $1.50 / $7.50. There is no free tier.
The featured image for this article was created with the help of AI.