Google announces Gemini 4 Argon with 1 million output token limit
Google says Gemini 4 Argon is rolling out soon, a new AI model built to handle tough jobs like coding, research, and cybersecurity.
With a massive 1 million output token limit (way up from before), it can take on bigger projects and more detailed tasks.
It's already showing off with top benchmark scores.
Gemini 4 Argon ties CWE-bench v1
Gemini 4 Argon shines in cybersecurity tests, tying for first place on CWE-bench v1 with a 68% score.
Google has baked in extra protections to guard against misuse and hacking tricks like prompt injection.
Gemini 4 Argon is rolling out soon for Google AI Ultra subscribers and paid API users (starting at $2 per million input tokens and $10 per million output tokens), plus Google is using it internally to boost data center efficiency and manage huge code migrations safely.