Gemini 4 Argon
Google DeepMind's new frontier model for long-horizon coding, legal and finance work, and cyber defence.
Quick verdict
- Best for
- Teams building long-running coding or research agents
- Pricing
- Introductory $2 / $10 per million tokens (standard $4 / $20); cached input $0.10
- Not ideal if
- Not generally available at launch — gated to the Fairwind Program with no public date
What is Gemini 4 Argon?
Gemini 4 Argon is Google's frontier model announced on September 30, 2026, built for long-running agentic work: real-world software engineering, enterprise legal and finance tasks, and cybersecurity defence. It accepts text, image, video, and speech, has a 1M-token context, and raises the output limit to 1M tokens so a single run can keep working for hundreds of thousands of tokens. At launch it is gated to vetted defenders through Google's Fairwind Program, with paid API customers and AI Ultra subscribers next.
Key features
- 1M-token context and a 1M-token output limit
- Long Decode Continuation to pause and resume extended responses
- Text, image, video, and speech input
- Google-reported leads on legal, finance, and software-engineering agent benchmarks
- Fairwind Program access for cyber defenders first
Pros
- Output limit an order of magnitude above most frontier models — suits very long agent runs
- Introductory price matches the $2/$10 tier set by Sonnet 5.5 and GPT-6.1 Sol
- Multimodal input in one model
Cons
- Not generally available at launch — gated to the Fairwind Program with no public date
- Introductory price doubles after an unannounced promotional period
- Benchmarks are Google-reported and unverified independently
Best for
Read more
Related comparisons
Alternatives to Gemini 4 Argon
GPT-6.1 Sol
OpenAI's efficiency model: near-Astra results at one-fifth of the token price.
Claude
Anthropic's AI — preferred by writers, developers, and researchers.
Gemini
Google's AI — best if you live in Gmail, Docs, and Search.