Google made Gemini 4 Argon generally available on 13 October. Developers and enterprises can reach it through the API in AI Studio and the Gemini platform, Google AI Ultra subscribers get it in the Antigravity coding tool, and Ultra and Pro subscribers have it in the Gemini app and AI Mode in Search. It had first been shown on 30 September to a limited group of trusted cybersecurity partners, so this widens access rather than announcing the model for the first time. Google also began expanding to hundreds more partners in its Fairwind cyber-defence programme.
The API model is called gemini-4-argon and has a unified window of about one million tokens covering input and output. Tools include code execution, function calling, file search and URL context. Introductory pricing is $2 per million input tokens and $10 per million output, with cached input 95% cheaper, and from 30 November it rises to $4 and $20. Google's pitch is deep multi-step work in coding, finance, legal, cybersecurity and science, in answers that hedge less.
“Gemini 4 Argon is just the beginning. We’re looking forward to bringing you the rest of the Gemini 4 family.”
Koray Kavukcuoglu, SVP Google DeepMind and Chief AI Architect, Google (author of the post), in Google's announcement · blog.google ↗
The performance claims are Google's. It says Argon ranks first on the Vals Index and scores 87% on the ERQA embodied-reasoning benchmark. Inside the company, using it in Antigravity raised agentic code submissions by over 35% and cut code rollbacks by nearly a third. Let's Data Science stresses that the figures are not independent and that the model card admits hallucinations and occasional slowness or timeouts.
Google offered a set of demonstrations. The model built 3D scenes in Blender, a racing game from photos of a desk, and a reconstruction of Chaco Canyon from about 50GB of scans. Apptronik's Apollo humanoid sank seven cornhole shots after learning the task from a video. Isomorphic Labs is using it in drug-discovery research, and Wiz applies it in its Scan for Good initiative.
There is a programme for researchers too: 10,000 scientists can get Antigravity's top usage tier free for six months, on a rolling basis, with applications open.
Google positions Argon for sustained, multi-step work in coding, finance, legal, cybersecurity and science. Its own use in Antigravity is the evidence it leans on most, and that evidence is self-reported too.
The pricing change is worth noting for anyone planning to build on it. The $2 and $10 rates apply only until 30 November, after which both input and output prices double. Let's Data Science also flags that the benchmark figures come from Google, so the cost and quality comparisons with rival models are still for outside testers to make.
Argon is the first member of a family. Google has given no dates for the rest of Gemini 4.
Sources · 2