Skip to content
  • Google
  • Gemini
  • Models

Google Gemini 4 Argon: what it is, price and availability

Gemini 4 Argon is Google's new frontier model, broadly available since 13 Oct 2026. Plans, $2/$10 intro API price, scores, and how it compares to Claude.

Quick answer: "Argon" is Gemini 4 Argon, Google's new frontier model. Google DeepMind announced it on 30 September 2026 for a small group of cyber defenders, and on 13 October 2026 made it broadly available in the Gemini API, AI Studio, Antigravity, the Gemini app and AI Mode in Search. The API model ID is gemini-4-argon. It costs $2 per million input tokens and $10 per million output tokens until 30 November, then $4 / $20.

Key facts

  • Announced 30 September 2026, broadly available 13 October 2026. Both posts are by Koray Kavukcuoglu of Google DeepMind.
  • API: gemini-4-argon on the paid tier of the Gemini API. It isn't on the free tier.
  • Price: $2 / $10 per million input/output tokens through 30 November 2026, then $4 / $20. Cached input is 95% off, and batch is half price.
  • Consumer access: Google AI Pro and Ultra in the Gemini app and AI Mode in Search. Antigravity, Google's coding agent app, needs Ultra.
  • Google says the API supports a 1M unified input-output token limit.
  • Independent scores: #1 on the Vals Index (68.90%), and 53 on the Artificial Analysis Intelligence Index, below Claude Opus 5.5 (58) and Sonnet 5.5 (56).

What Google says Argon is for

In the launch post, Google describes Argon as a model that keeps reasoning through long, complicated workflows, aimed at software engineering, enterprise work like legal and finance, and cybersecurity defense. It started with a limited release through its Fairwind Program for trusted cyber defenders, saying a phased approach was needed for capabilities at this level. With the 13 October post, the model went broadly available, and Google said it's expanding Fairwind to hundreds more partners, who can get a version without cyber guardrails.

Argon works longer and harder to solve highly complex scientific, coding, and agentic problems.

— Koray Kavukcuoglu, Google DeepMind · Read the original

Where you can use it

  • Gemini API, paid tier, in Google AI Studio and Android Studio. AI Studio's pricing page says Argon isn't on the free tier.
  • Gemini for business and the Gemini platform API for enterprises.
  • Google AI Ultra: Antigravity, both the desktop app (macOS, Windows, Linux) and the terminal CLI, plus the Gemini app and AI Mode in Search.
  • Google AI Pro: the Gemini app and AI Mode in Search, but not Antigravity.
  • Researchers can apply for six months of free access to Antigravity's highest usage tier, subject to selection.

Pricing

Per Google's AI Studio pricing page, Argon costs $2 per million input tokens and $10 per million output tokens (thinking tokens count as output) through 30 November 2026. From 1 December it's $4 / $20. Implicit context caching is $0.10 per million tokens now and $0.20 later, and batch requests are half price. There's no separate surcharge tier for long prompts. For comparison, Anthropic lists Claude Sonnet 5.5 at $2 / $10 and Opus 5.5 at $4 / $20, so Argon's intro price matches Sonnet 5.5 and its standard price matches Opus 5.5 per token. What you actually pay per task also depends on how many tokens each model uses.

Benchmarks: Google's numbers and independent ones

Google's own figures, from its posts: 77.9% on DeepSWE v1.1 for long-horizon software engineering, 51.3% on Zapier's AutomationBench, 87% on the ERQA embodied-reasoning test, and a 0.7% attack success rate on the Gray Swan indirect prompt-injection benchmark across 15 adaptive attempts. VentureBeat notes that in Google's own comparison table, Argon trails Claude Opus 5.5 on Terminal-Bench 4.0 (57.4% against 66.4%).

Independent results are mixed. Vals AI ranks Argon #1 on its Vals Index at 68.90%, just ahead of Claude Sonnet 5.5 (67.04%), Opus 5.5 (66.97%) and Fable 5.1 (65.83%). Artificial Analysis gives Argon (High) 53 on its Intelligence Index v4.3.2, at $1.99 per index task, level with Claude Fable 5.1 and GPT-6 Astra but below Opus 5.5 (58) and Sonnet 5.5 (56), as of 13 October 2026. Scores change as labs re-run evaluations.

How it relates to Codex and Claude Code

Google's equivalent of Codex and Claude Code is Antigravity, an agent-first development app with a desktop app and a terminal CLI. In May, Google announced that Gemini CLI would be replaced by Antigravity CLI, and it stopped serving Google AI Pro and Ultra subscribers on 18 June 2026. Argon in Antigravity needs Google AI Ultra. On a Pro plan you can chat with Argon in the Gemini app, but you can't use it as a coding agent there. If you use a third-party coding tool with your own API key, gemini-4-argon is available on the paid API tier.

Sources

  1. Gemini 4 Argon: our next era of frontier intelligence · Google ·
  2. Gemini 4 Argon: Built to solve your hardest problems · Google ·
  3. Gemini 4 Argon is now available for developers · Google ·
  4. Gemini API pricing · Google AI Studio
  5. Bringing Gemini 4 Argon to Android developers · Android Developers Blog ·
  6. Transitioning Gemini CLI to Antigravity CLI · Google Developers Blog ·
  7. Fairwind Program · Google DeepMind
  8. Gemini 4 Argon on the Vals Index · Vals AI
  9. Gemini 4 Argon (High) · Artificial Analysis
  10. Google makes Gemini 4 Argon broadly available · VentureBeat ·
  11. Models overview (Claude prices) · Anthropic Docs

Retro AI Tracker is an independent fan site. Quotes are short excerpts linked to their originals. See how we verify or report a correction.

Computer mode

Prefer a retro PC?

Boot the MYPC98 desktop: windows, the big reset button, the community inbox, Token Hop and lo-fi radio. The website stays the default.

Boot computer mode