Google began rolling out Gemini 4 Argon, its latest frontier model, on October 13, and the access rules are more restrictive than the headline “available for everyone” suggests. Argon is open to developers and enterprises through the API, to Google AI Pro and Ultra subscribers in the Gemini app and AI Mode in Search, and only to AI Ultra subscribers in Google Antigravity and AI Studio’s coding workflows. The API price Google is advertising is temporary.

Who gets Argon, and where

Google’s launch post says Argon is available through the API in AI Studio and on the Gemini platform, for Google AI Ultra subscribers in Antigravity, and for AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search. Ars Technica adds the plan prices and a caveat: AI Pro costs $20 a month, AI Ultra starts at $99 a month, and free users do not get Argon. Google had also recently restricted free access to its Flash models, leaving free users with Flash-Lite only.

For coding-focused products the gates are tighter. Ars reports that in Antigravity and AI Studio the model is limited to Ultra plans, while Pro subscribers can still generate code with Argon in the Gemini app’s Canvas mode. Google has not said when, or whether, Argon will reach Antigravity on lower tiers. Gemini 3.1 Pro remains available for now for people who want something more capable than a Flash model.

The API price clock

The headline number is an “introductory price” of $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%. According to Ars, that deal runs through November 30. After that date the rate rises to $4 per million input tokens and $20 per million output tokens, double the launch price. Anyone planning to build a workflow around Argon should model costs at the later price, since a budget that works at $2/$10 will not necessarily work at $4/$20.

A worked example shows why. A pipeline that sends 50 million input tokens and receives 5 million output tokens in a month would cost $150 at the introductory rate ($100 for input plus $50 for output) and $300 afterwards, before any caching savings. Heavy agentic coding runs consume many more tokens than that, which is likely why Google reserves the multi-threaded workflows in Antigravity for the top plan.

What Google claims Argon can do

Google positions Argon for “deep work” in coding, finance, legal, cybersecurity and science, and says it ranks first on the Vals Index, a benchmark of performance on economically relevant tasks. These are Google’s own statements, and we have not seen an independent replication. The company also shared internal examples: it says Argon in Antigravity helped teams raise agentic code submissions by more than 35% while cutting code rollbacks by nearly a third, helped the Chrome team with a refactor across hundreds of files, and made Google’s open Gemma models 2.85 times faster in an inference setup on an A100 GPU.

Ars reports Google originally said it needed to move cautiously because of the potential for misuse, and describes safeguards including training to refuse malicious instructions and an automated misalignment monitor that can stop processing if it detects an unintended workflow. Ars notes that nothing about the model’s capabilities has changed since the benchmarks Google published late last month.

Why it matters

Argon illustrates a pattern that is now common: the most capable model arrives first on the most expensive plan, and price protections are time-limited. For individuals, the practical decision is whether Argon’s gains justify moving from the free tier to a $20 plan. For teams, it is whether to commit to an API price that may double in six weeks. Our guide to how AI pricing works explains the units involved, and how to cut your AI API bill covers caching and routing that can offset a price rise.

What to watch next

  • Whether Google extends Argon to Pro subscribers in Antigravity and AI Studio.
  • Independent evaluations of Argon on public leaderboards, not only Google’s own results.
  • Any change to the November 30 price step or the cached-token discount.
  • How rivals price their next frontier releases in response.
  • Whether Gemini 3.1 Pro stays available once Argon is widely rolled out.

Key takeaways

  • Argon is available to paying users and API customers, not to free users.
  • AI Pro ($20/month) gets it in the Gemini app; Antigravity and AI Studio coding require AI Ultra (from $99/month).
  • The $2/$10 per million token API price ends November 30 and becomes $4/$20.
  • Cached input tokens are discounted 95%.
  • Performance claims, including the Vals Index ranking, come from Google.

Frequently asked questions

Can I use Gemini 4 Argon for free?

Not at the moment. Google’s rollout covers paid plans and the paid API, and free Gemini users do not have access.

Will the API price really double?

According to Ars Technica’s reading of Google’s terms, yes: the introductory $2/$10 pricing ends on November 30, after which the price is $4 per million input tokens and $20 per million output tokens.

Should I switch my production workload to Argon now?

Test it, but budget for the post-November price and compare results against cheaper models on your own tasks before committing.