Gemini 4 Argon Opens to Pro Subscribers and Lands on Vercel AI Gateway
Google opens Gemini 4 Argon to Pro and Ultra plans, publishes it on Vercel AI Gateway the same day, and sets an introductory price of US$2 per million tokens.
Google opened Gemini 4 Argon to AI Pro and AI Ultra subscribers in the Gemini app and in Search's AI mode on October 13, 2026, the same day Vercel published the model on its AI Gateway under the key google/gemini-4-argon.
Argon was introduced in late September with restricted access to the Fairwind program, made up of governments and trusted cybersecurity partners, according to Google's official blog. The model card documents a unified 1 million token window and an introductory price of US$2 per million input tokens and US$10 per million output tokens, valid until November 30, 2026. Engadget reported on October 13 that the model is already invoked from the Gemini API and the Antigravity agentic development platform. For a product team in Mexico, the real shift is not the benchmark: it is that the model appears in a third-party gateway catalog on launch day.
9to5Google specified that AI Pro and AI Ultra access the model in the app and in AI Mode, and that Antigravity, AI Studio and Android Studio Canary reserve it for the Ultra plan. Google also opened 10,000 free spots at the highest tier of Antigravity for six months for researchers. On the operational side, ecosistemastartup details that the gateway exposes the model via AI SDK and the Chat Completions, Responses and Anthropic Messages APIs, and that it replaces thinking budgets with low, medium and high reasoning levels configurable with the reasoning parameter. The 1 million token window allows loading an entire repository or several spreadsheets without fragmenting the context. Vercel does not charge a markup over the provider's price for retries or routing rules, and offers Zero Data Retention, a frequent requirement in corporate due diligence.
Price per token does not decide cost per task: the introductory rate doubles on December 1, 2026. What remains on the table is measuring the team's real spend before that date and defining a backup model on the gateway.
This note was drafted with the assistance of artificial intelligence from verified sources and reviewed by a human editor before publication.