Google DeepMind Launches Gemini 4 Argon with 1M Output
Google DeepMind introduced Gemini 4 Argon, a new frontier model aimed at coding, enterprise knowledge work, and cyber defense. It is currently rolling out to a limited set of trusted testers through the Fairwind Program, including government users and cyber defenders. Argon demonstrates SOTA results in 13 of 19 credible benchmarks, and offers an industry-first 1M-token output via Long Decode Continuation. Pricing is set at $4/$20 per million input/output tokens, with a 50% introductory discount and 95% discount for cached input. Google also claims internal agentic deployments have freed over 300 TiB of memory and are migrating 800K lines of C/C++ code to Rust. Early evaluations show Argon matching GPT-6 Astra on the Artificial Analysis Intelligence Index (53), but also show higher token usage per task and trade-offs in accuracy vs. hallucination rates.
- •Google DeepMind unveiled Gemini 4 Argon, a frontier model for coding, enterprise knowledge, and cybersecurity, rolling out to trusted testers via the Fairwind Program.
- •Argon sets a new industry record with 1M output tokens via the Long Decode Continuation API feature, up from the previous 64K limit.
- •Standard pricing is $4/$20 per million input/output tokens, with a 50% introductory discount and a 95% discount for cached input.
