4 hrs ago
Google Unveils Gemini 4 Argon With One-Million-Token Output
Google has made a new AI model called Gemini 4 Argon.
It can write much longer answers than earlier versions.
Its limit is one million small pieces of text, called tokens.
This may help it work on large software projects, long documents and complicated research.
Google says its employees are already using it for coding and engineering tasks.
The company also says Argon can find and fix some computer security problems.
Because that power could be misused, Google is first giving it to trusted cybersecurity experts.
Later, Google plans to offer it to developers, businesses and some subscribers.
Google has not announced the final date for wider consumer access.
Google says Gemini 4 Argon can generate up to one million tokens, compared with the previous 64,000-token limit.
The model is being used internally for coding, research, debugging and large-scale engineering projects.
Google reported that Argon converted 32,000 lines of SIMD code to Rust, producing a version 2.7 times faster than an earlier Rust port.
The model is initially being tested by trusted cyber defenders because of its ability to identify and patch software vulnerabilities.
Access will expand to paid API customers and Google AI Ultra subscribers, with initial pricing of $2 per million input tokens and $10 per million output tokens.
- Who
- Google unveiled Gemini 4 Argon and said its employees and selected cyber defenders are testing it.
- What
- A frontier AI model that can generate up to one million tokens in one output and handle long, complex workflows.
- Where
- The model is being used internally at Google and tested through its Fairwind Program.
- When
- The articles do not provide a specific calendar date; Google has announced the model and begun limited testing.
- Why
- Google says the model is designed for extended coding, research, enterprise work and cybersecurity tasks, while limited access allows additional safety testing.
Key facts
- Model
- Gemini 4 Argon
- Maximum output
- One million tokens
- Previous limit
- 64,000 tokens
- Internal uses
- Coding, debugging, research, quantum-computing optimization, data-center memory efficiency and code migration
- Initial access
- Selected trusted cyber defenders through Google's Fairwind Program
- Initial pricing
- $2 per million input tokens and $10 per million output tokens
- Reported benchmark results
- 77.9% on DeepSWE v1.1 and 51.3% on Zapier's AutomationBench
Sources
Related news
Google Unveils Gemini 4 Argon for Long-Horizon AI Workflows
Google Launches Gemini 4 Argon for Cybersecurity and Coding
Google’s Gemini 4 Argon targets coding, finance and legal work
Google Announces Gemini 4 Flagship Argon After Months Of Delays
Google Restricts Gemini 4 Argon Access Amid AI Safety Concerns
SpaceX AI Unveils Faster, Cheaper Grok 4.7 Coding Model
