Google opened Gemini 4 Argon to broader access on October 13. Developers and enterprises can use the API, and Google AI Pro and Ultra subscribers can use it in the Gemini app and AI Mode in Search. The release extends access beyond Google’s early partners, though each product has its own access terms.
For builders, the main change is that Argon is now available for paid API work. Google lists an introductory price of $2 per million input tokens and $10 per million output tokens, with a 95% discount on cached input.
Which tier opens which product
Google describes the rollout as global. Its announcement lists these routes:
| Use | Access Google lists |
|---|---|
| Build through the Gemini API | Paid tier in AI Studio and Android Studio |
| Use Argon through a personal Ultra subscription | Google Antigravity and AI Studio |
| Enterprise work | Gemini for business, the API and Google Antigravity |
| Chat and Search | Google AI Pro or Ultra in the Gemini app and AI Mode |
| Scientific research | Applications for six months of Antigravity’s highest usage tier at no charge |
| Cybersecurity defense | Applications to the Fairwind Program for access without cyber guardrails |
Google says it will offer the research program to 10,000 scientists, with rolling onboarding subject to eligibility, capacity and selection. It is also beginning an expansion to hundreds of additional Fairwind partners. Those programs require separate applications.
The API has a one-million-token unified input-output limit, according to the announcement. Google directs developers to its continuation protocol documentation for setup. Teams planning long runs should account for both the prompt and the response within that stated limit.
How Argon compares with GPT-6 and Opus
The DeepMind model page publishes a comparison against GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5. Selected results against Astra and Opus show why a single ranking gives builders too little detail:
| Benchmark | Gemini 4 Argon | GPT-6 Astra | Claude Opus 5.5 |
|---|---|---|---|
| Vals Index | 68.9% | 63.1% | 67.0% |
| AutomationBench | 51.3% | 41.4% | 42.5% |
| DeepSWE v1.1 | 77.9% | 74.1% | 74.2% |
| FrontierSWE v2 | 55.0% | 65.5% | 62.3% |
| Terminal-bench 4.0 | 57.4% | 58.2% | 66.4% |
| GraphWalks, 256k to 1M, BFS F1 | 84.2% | 71.8% | 66.8% |
| CWE-bench v1 | 68.0% | 68.0% | 67.0% |
These are Google’s published results. Argon leads these rivals on the Vals knowledge-work index, AutomationBench and DeepSWE. Astra leads on FrontierSWE, while Opus leads on Terminal-bench. Argon ties Astra on the cybersecurity benchmark.
Google also reports engineering gains inside the company. It says Argon in Antigravity helped teams increase agentic code submissions by more than 35% and reduce rollbacks by nearly a third. Those internal results give a separate example of deployment, alongside the benchmark scores.
Sources
- Google: Gemini 4 Argon: Built to solve your hardest problems. Release date, access tiers, introductory pricing and deployment claims. Lead artwork credited to Google.
- Google DeepMind: Gemini. Model comparison table and research and security program details.
