Gemini 4 Argon: price, benchmarks, access

What will Gemini 4 Argon cost you?

Introductory pricing runs through November 30, 2026. Pick your billing date and usage, and the calculator shows what you pay now, what you pay after the hike, and how that compares to rival models.

One more caveat before you trust the math: cheaper per token is not always cheaper per task. One independent test ran ten identical sessions and found Argon wrote about 3.7× more tokens than GPT-6 Astra, which would erase the per-token advantage at the post-November price. Source: independent 10-session cost test (Oct 2026).

What is Gemini 4 Argon?

Argon is the first model of the Gemini 4 generation and Google's first new flagship since Gemini 3 arrived in late 2025. Anthropic and OpenAI shipped strong models in the months between, so this launch is Google's attempt to get back in the frontier race. (GenAInews)

Google's positioning word is "deep work": long-horizon reasoning and agentic tasks rather than short chat answers. The model targets software engineering, enterprise knowledge work (legal, finance, tax), cybersecurity defense, and science.

The headline spec: 1M output tokens

The biggest technical change is output length. Argon can generate up to 1 million tokens in a single response, up from 64,000 on earlier Gemini models. That relies on a Gemini API feature called Long Decode Continuation: it pauses long responses and resumes them in later calls so requests don't time out. For developers, that means large refactors or long reports without splitting work across turns. (deeplearning.ai)

The context window is a unified 1 million tokens, shared across input, thinking, and output. Reasoning is tunable across three levels (low, medium, high), so you can trade depth for speed and cost per request. Input accepts text, images, video, and audio; output is text. (dev.to explainer)

What Google claims it does better

• Answers more directly, with less hedging and repetition than earlier models. (The AI Rankings)
• Leads on knowledge work that rewards precision: legal research with citation fidelity, financial analysis, cybersecurity defense.
• Already runs inside Google: it patches security vulnerabilities, halved the compute behind YouTube comments, helped debug the physics simulator for Wing drones, and produced a Rust port of the libgav1 video decoder that runs 2.7× faster than the existing one. (GenAInews)
• In Antigravity, Google's coding tool, teams using Argon submitted over 35% more agentic code with nearly a third fewer rollbacks. (Reuters)

Benchmarks: what the scores actually say

Independent: Artificial Analysis Intelligence Index

Artificial Analysis scored Argon 53 on the Intelligence Index (v4.3.2) at high reasoning, tied with GPT-6 Astra and Fable 5.1, five points below Claude Opus 5.5 (58). At the introductory price, Artificial Analysis puts Argon's cost at about $1.99 per index task, roughly 60% of GPT-6 Astra's. (The AI Rankings; Cryptopolitan)

Google's own numbers (vendor-claimed, not independently verified)

Sources: deeplearning.ai, Gadgets360. Google has not disclosed parameter count, architecture, latency, or knowledge cutoff.

Where you can use it

Early enterprise testers include Deloitte, LexisNexis, Ambience Healthcare, and Isomorphic Labs. Google is also giving 10,000 scientists six months of free access to Antigravity's highest usage tier. (Reuters)

Availability confirmed at global rollout on October 13, 2026, per Google and Reuters.

How we got here

• September 30, 2026: Google DeepMind announces Gemini 4 Argon. Access is limited to trusted cyber defenders through the Fairwind Program. (MarkTechPost)
• October 13, 2026: Global rollout to developers, enterprises, and Pro/Ultra subscribers. Sundar Pichai confirms on X. (Reuters)
• November 30, 2026: Introductory $2/$10 pricing ends.
• December 1, 2026: Standard pricing begins: $4 per million input tokens, $20 per million output.

Context: the launch follows a rough stretch: Google scrapped the planned Gemini 3.5 Pro and overhauled DeepMind leadership after co-founder and CEO Demis Hassabis stepped aside. (Reuters)

Sources

• Reuters · Google launches flagship AI model Gemini 4 Argon (Oct 13, 2026)
• deeplearning.ai · Argon: a low-hallucination model built for knowledge work
• The AI Rankings · Gemini 4 Argon: benchmarks, pricing and access
• Gadgets360 · Google announces Gemini 4 Argon
• GenAInews · Google releases Gemini 4 Argon
• Cryptopolitan · pricing vs OpenAI and Anthropic
• MarkTechPost · 1M output tokens announcement (Sep 30, 2026)
• Independent 10-session cost test vs Opus 5.5, GPT-6 Astra, Muse Wonder (YouTube)

Page published 2026-10-15. Prices and availability change; the calculator notes which inputs are assumptions.