Google on Tuesday, Oct. 13, 2026 opened Gemini 4 Argon, its most capable AI model, to paying users. It is now in the Gemini app and AI Mode in Search for Google AI Pro and AI Ultra subscribers, in the paid tier of the Gemini API for developers and enterprises, and in Google's Antigravity coding tool for AI Ultra subscribers, Google DeepMind's Koray Kavukcuoglu wrote in Google's announcement. 9to5Google reports Argon was announced at the end of September, when access was limited.

Who gets Argon, and what it costs

  • Gemini app and AI Mode: Google AI Pro and AI Ultra subscribers. 9to5Google reports the model picker for those tiers now shows 3.5 Flash-Lite, 3.8 Flash, 3.1 Pro and 4 Argon.
  • Antigravity, AI Studio and Android Studio (Canary): AI Ultra required, per 9to5Google.
  • Gemini API: paid tier only; Google's pricing page says Argon is not available on the free tier. The introductory price is $2 per million input tokens and $10 per million output tokens through Nov. 30, 2026, rising to $4 and $20 from Dec. 1. Cached input is 95% off.
  • Context: Google says Argon has a 1 million token unified input-output limit in the Gemini API.

Google is also offering 10,000 scientists its highest Antigravity usage tier free for six months, subject to eligibility and capacity, and is expanding access to hundreds more partners in its Fairwind Program, which gives vetted cyber defenders a version of Argon without cyber guardrails.

What Google claims for it

Google says Argon ranks first on the Vals Index, which measures performance on finance, coding, legal and tax work, and that inside Google, Argon in Antigravity raised agentic code submissions by more than 35% while cutting code rollbacks by nearly a third. These are Google's own figures.

On safety, Google says external biology red-teamers could unlock less than 1% of the bio sub-tasks it tests for, that shortcut attempts on coding benchmarks fell to 2%, and that it ships an automated misalignment monitor that reads the agent's reasoning and actions to block unintended workarounds in real time. It also says Argon cut prompt-injection attack success to 0.7% on the Gray Swan benchmark.

The robotics angle

Google says Argon scored 87% on the Embodied Reasoning (ERQA) benchmark for physical and spatial problems, ahead of other frontier models. In an internal test, Argon watched a video of a person playing cornhole, built the simulation code itself, and the skill was transferred to Apptronik's Apollo humanoid, which sank seven shots in a row in a real-world trial, Google says. Apptronik is one of the humanoid makers competing in the same field as Tesla's Optimus; see our Optimus tracker for Tesla's own progress.

The release lands the same week xAI shipped Grok Voice 3.0 and rolled Grok into new combined X subscription plans.