Google has introduced Gemini 4 Argon, its latest frontier AI model designed for demanding tasks such as software engineering, knowledge work, and cybersecurity.
The company is initially giving access to trusted members of its Fairwind Program, allowing selected users to test Argon for software vulnerabilities and fix potential issues before a broader rollout. Google is also participating in the US government’s voluntary pre-release testing program for advanced AI models.
Google is positioning Argon aggressively on price. The model will launch at introductory rates of $2 per million input tokens and $10 per million output tokens.
That makes it considerably cheaper than several competing flagship models. Anthropic charges $10 per million input tokens and $50 per million output tokens for Fable 5.1, while OpenAI’s GPT-6 Astra carries similar pricing. OpenAI’s newer GPT-6.1 Sol matches Argon’s $2 input and $10 output pricing.
Gemini 4 Argon represents Google’s latest attempt to strengthen its position at the frontier of AI after intense competition from OpenAI and Anthropic.
Google says Argon is already being used internally for coding, research, writing, and other workflows. However, Bloomberg reported that opinions inside the company differ over whether the model can fully compete with the strongest systems from OpenAI and Anthropic.
Google plans to release Gemini 4 Argon gradually before expanding access to enterprise customers and the wider market.
Eventually, the company expects Argon to power many of its products and services, following the strategy it used with Gemini 3 across platforms including Google Search and YouTube.