Google says Gemini 4 Argon is now globally available via the API in AI Studio and the Gemini platform, plus in Antigravity for Ultra subscribers and in the Gemini app and AI Mode in Search for Ultra and Pro subscribers.
Update, Oct 19, 16:52 UTC · via Deepmind
Deepmind reported that the post dated Oct 13, 2026 was authored by Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, Google, and described the pricing as an "introductory price". The outlet listed additional focus areas including work "To find and patch critical cybersecurity vulnerabilities" and "To enable novel breakthroughs in quantum computing, nuclear fusion, and life sciences".
Deepmind reported that the Chrome refactor was done by "Piping multiple clean code changes at a time" and "completing months of work in weeks", enabling pages to move "between tabs, side panels, and floating windows without reloading" while "eliminating an entire category of browser crashes". For the ADK optimization, the team gave Argon their ".agents/skills rulebook" and a "one-paragraph goal to eliminate framework overhead without breaking existing apps" with changes now live in "google/adk-python", and for xTB the work is now "1.8x faster at conformer search" with changes "contributed to the official xTB repository on GitHub".
I read this as a price-led land grab for deep work, not just a model drop. The Gemini 4 Argon API pricing is set at $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off, and Google claims Argon ranks #1 on the Vals Index across finance, coding, legal and tax. Per the Gemini API pricing page, the model is not available on the Free tier.
That is not generous, it is strategic.
Google is using its own hardest jobs as proof. The company says Argon in Antigravity increased agentic code submissions by more than 35% while reducing rollbacks by nearly a third, helped untangle a ~6 million line Chrome folder inside a 40 million line Chromium codebase to let pages move between tabs without reloads, and delivered overnight optimizations to its Agent Development Kit that cut in-memory session prep time by 40% and cold imports by 21%. It also cites a 2.85x inference speedup on an A100 via vLLM for Gemma 4 E2B, from ~37,900 to over 108,000 tokens per second with 100% token parity, and up to 2.3x faster molecular structure prediction in the open-source xTB package.
The squeeze lands on inference. If frontier reasoning is now $2 to get in and almost free to stay in via caching, startups selling speed and cost per token have to prove they beat Google on latency, throughput, or private deployment, not just price. StartupHub.ai data as of this month shows AI hardware and inference optimization startups still carry strong traction for private and edge deployment, even as hyperscaler API pricing compresses gross margins for pure token arbitrage.
