Is Gemini 4 out?
Gemini 4 is out, and it's three models, not one. Google shipped Argon first, then Carbon, then Hydrogen, and all three were live by 12 November 2026. Each has its own plan, its own price and its own job. The chart dates every step, mostly from Google's own posts.

Dates for Argon, Carbon's developer release and Hydrogen come from Google's blog posts, listed under sources. Carbon's 28 October entry is the release date Artificial Analysis prints for it. Google's own Carbon post is the 10 November one. Here are the three side by side as of 15 November 2026.
| As of 15 November 2026 | Hydrogen | Argon | Carbon |
|---|---|---|---|
| Google's description | Everyday coding and agentic work | A more efficient frontier reasoning model | The strongest reasoning model in the family |
| Developers get it | 12 Nov 2026 | 13 Oct 2026 | 10 Nov 2026 |
| Model code | gemini-4-hydrogen | gemini-4-argon | gemini-4-carbon |
| Context window | 1M tokens | 1M tokens | 1M tokens |
| Gemini app plans | Plus, Pro, Ultra | Pro, Ultra | Ultra |
| API price per 1M tokens, input and output | $1.50 and $7.50 | $2 and $10, then $4 and $20 | $2 and $10, then $4 and $20 |
| Free API tier | Yes | No | No |
| Artificial Analysis index, top setting | 54 | 50 | 60 |
The descriptions and model codes are from Google's models page. The prices are from its pricing page. The index scores are the ones Artificial Analysis lists for each family, which is a benchmark site's measurement and not Google's.
What has Google said about each Gemini 4 model?
Argon came first. Google's 30 September post put it in the hands of a set of trusted cyber defenders through its Fairwind Program. It said the wider release would start "with paid API customers and Google AI Ultra subscribers" and gave no date. It also raised the output limit to 1M tokens, up from 64K. The 13 October post called Argon "now globally available to developers and enterprises via the Gemini API" and added the Gemini app, AI Mode and Antigravity for subscribers.
Argon's model card, also dated 13 October, gives a March 2026 knowledge cutoff. It accepts text, images, audio and video, and returns text. It carries a warning too: "in some domains the model's knowledge may be less up to date."
Carbon was next. Google's 10 November developer post calls it "our strongest model with frontier performance in cybersecurity and long-horizon capabilities." Developers can use it on the paid API tier in Google AI Studio and Android Studio. Ultra subscribers get it in AI Studio and Antigravity. The post prints no price and no benchmark scores.
Hydrogen is the newest. Koray Kavukcuoglu, Google's chief AI architect, announced it on 12 November. Google says it finishes tasks 2.5 times faster than Gemini 3.8 Flash at its default setting. It scores 57% on Terminal-Bench 4.1, against 18.2% for Flash, by Google's count. Hydrogen adds a fourth thinking level, xhigh. A companion post says "Hydrogen is now available globally in Google Search for all users."
How much does Gemini 4 cost?
Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, and Argon and Carbon cost $2 and $10 until 30 November 2026. On 1 December both go to $4 and $20. Google's pricing page lists the same rates for Argon and Carbon. A standard answer here is 1,000 tokens in and 500 out.
| API, standard tier, as of 15 November 2026 | Input per 1M | Output per 1M | One standard answer |
|---|---|---|---|
| Hydrogen | $1.50 | $7.50 | $0.00525 |
| Argon or Carbon, through 30 Nov | $2.00 | $10.00 | $0.007 |
| Argon or Carbon, from 1 Dec | $4.00 | $20.00 | $0.014 |
The last column is our arithmetic from Google's rates, not a Google figure. It shows the December gap: Argon or Carbon will cost about 2.7 times what Hydrogen does for the same answer. Batch and Flex are half the standard rate, and Priority runs 80% higher. Cached input is cheaper still. Google lists Argon and Carbon as unavailable on the Free tier. Hydrogen has one, with a catch: Google says free-tier data is used to improve its products.
The API isn't the Gemini app. With the API you pay for what you use, and with a chat plan you pay a flat monthly fee, whatever you send. Gemini API pricing covers the older models and the rate limits.
Which Gemini 4 model do you get on each Google AI plan?
Free and Plus get Hydrogen, Pro adds Argon, and only Ultra gets Carbon in the Gemini app. Google's plans page says Plus "includes Gemini 4 Hydrogen," Pro "includes Gemini 4 Argon" and Ultra "includes Gemini 4 Carbon." Its comparison table adds that both Ultra tiers also reach Argon. The context window and usage figures come from Google's limits page.
| Plan, US price | Gemini 4 models | Context in one chat | Usage |
|---|---|---|---|
| Free, $0 | Hydrogen through auto model selection | 32K tokens | Standard limits |
| Google AI Plus | Hydrogen | 128K tokens | 2x standard |
| Google AI Pro, $19.99 | Argon and Hydrogen | 1M tokens | 4x standard |
| Google AI Ultra, from $99.99 | Carbon, Argon and Hydrogen | 1M tokens | 5x or 20x Pro |
The free row needs a caution. The Hydrogen post says free users get it "through auto model selection" when it suits the prompt, and the limits page names no Gemini 4 model. Google's subscriptions page still lists a Flash-Lite model for Free. So you can't pick Hydrogen on Free. Google picks it for you, some of the time. Limits refresh every five hours up to a weekly cap.
The gap that costs money is Carbon. Google reserves the strongest model for the plan that starts at $99.99, and the API charges $2 per million input tokens for it. Whizi isn't Google's Gemini app, Antigravity or AI Mode. It's a separate chat app that calls the models. Its model list includes Gemini 4 Hydrogen, Argon and Carbon on the Powerhouse plan at $49.99 a month. Whizi Pro at $29.99 carries Gemini 3 models next to Claude and GPT, with a first week of $1.99. The free plan gives 10 messages a day on Whizi AI, Auto and Claude Haiku 5.5, no card.
How does Gemini 4 compare with GPT-6 and Claude?
Argon wins most of the benchmarks Google chose and loses on terminal work and some science tests. Google's Argon model card compares it with OpenAI's GPT-6 Astra and with Claude Opus 5.5 and Fable 5.1. These are Google's own runs, as of October 2026. Higher is better.
| Benchmark, Google's numbers | Argon | GPT-6 Astra | Claude Opus 5.5 | Claude Fable 5.1 |
|---|---|---|---|---|
| DeepSWE v1.1, coding | 77.9% | 74.1% | 74.2% | 67.4% |
| AutomationBench | 51.3% | 41.4% | 42.5% | 31.4% |
| GraphWalks 256k to 1M, long context | 84.2% | 71.8% | 66.8% | 65.0% |
| FrontierSWE v2, coding | 55.0% | 65.5% | 62.3% | 56.3% |
| Terminal-bench 4.0 | 57.4% | 58.2% | 66.4% | 57.9% |
| Terminal-Bench Science 0.1 | 57.6% | 68.1% | 63.3% | 52.6% |
Read the pattern, not the winner. Argon leads where the job is reading a huge input or running a business workflow. It trails where the job is deep command-line work or hard science. The card compares Argon only. Google's Carbon post prints no scores, and its Hydrogen post compares that model with other Gemini models, not with rivals. Gemini vs ChatGPT adds the blind-vote scores and the plan-by-plan comparison.
Which Gemini 4 model should you pick?
Pick Hydrogen for daily chat and coding, Argon if you pay for AI Pro and the task is long or messy, and Carbon only when a harder task has beaten the other two. Hydrogen is the model Google built for speed, and it's the only one you can try for nothing. Artificial Analysis scores it 54 at its top setting, above Argon's 50.
Argon is the model that Pro subscribers get, and its 1M context and its long-input scores suit big documents. Carbon is the top of the family at 60 on the same index. It costs the same as Argon through the API, so developers can try it without an Ultra plan. For a chat user, Carbon means $99.99 a month, so test Hydrogen and Argon on your own work first.
Is there a Gemini 4 Pro, Helix or Krypton?
Not according to Google. Its models page lists three Gemini 4 models, Carbon, Argon and Hydrogen, and no others. Autocomplete suggests names like Gemini 4 Pro, Helix and Krypton, but we found no Google post, model card or price page for any of them on 15 November 2026. If a site gives you details on one of them, you should treat that as a guess.
Google can add models, and it shipped three in six weeks. If you are about to pay for one of these models, check the models page for the current list first, because this page may be a few days behind it.
- Gemini 4 is three models: Hydrogen from 12 November, Carbon for developers from 10 November, Argon from 13 October
- Hydrogen is the only one with a free API tier, and it's in Google AI Plus, Pro and Ultra
- Argon and Carbon double to $4 and $20 per million tokens on 1 December 2026
- Carbon in the Gemini app needs Google AI Ultra, from $99.99 a month
- Check the model drop-down in your own Gemini app, since free users get Hydrogen only through auto model selection
Frequently asked questions
When was Gemini 4 released?
Google announced Gemini 4 Argon on 30 September 2026 to cyber defenders and opened it to developers and subscribers on 13 October. Carbon reached developers on 10 November and Hydrogen was released on 12 November.
Is Gemini 4 free?
Partly. Hydrogen has a free API tier and is in Google Search for all users. Free Gemini app users get it through auto model selection. Argon and Carbon aren't on the free API tier and need a paid Google AI plan in the app.
What is Gemini 4 Hydrogen?
Gemini 4 Hydrogen is Google's fast, lower-cost Gemini 4 model, released on 12 November 2026. It costs $1.50 per million input tokens and $7.50 per million output tokens, with a 1M-token context window, and Google AI Plus includes it.
Which Google AI plan includes Gemini 4 Carbon?
Google AI Ultra, which starts at $99.99 a month in the US. Carbon is also open to developers on the paid API tier, at $2 per million input tokens until 30 November 2026 and $4 after.
Is Gemini 4 Hydrogen better than Argon?
On the Artificial Analysis index Hydrogen scores 54 at its xhigh setting and 51 at high, against 50 for Argon. That is the site's measurement, not Google's. Hydrogen also costs less, at $1.50 per million input tokens against $2 for Argon until 30 November 2026.
How does Gemini 4 Hydrogen compare with Gemini 3.8 Flash?
Google says Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash at its high default setting. It also reports 57% on Terminal-Bench 4.1 against 18.2% for Flash. Both figures are Google's own, from its 12 November 2026 post.