Google DeepMind unveils Gemini 4 Argon

Google DeepMind has announced Gemini 4 Argon, its new frontier model. It is rolling out first to a set of trusted cyber defenders through the company's Fairwind Program. Google says it is taking part in the U.S. government's voluntary pre-release access process and will expand access gradually. The model will reach developers, enterprises and consumers "as soon as possible," after Google adjusts its guardrails. It has not given a date.

Introductory pricing is $2 per million input tokens and $10 per million output tokens. Cached input tokens cost 95% less than standard input. The output token limit rises to 1M, up from 64K on the previous model.

Google cites these benchmark results:

• DeepSWE v1.1: 77.9%, a new state of the art on long-horizon software engineering
• LVBench: 91.7% on long-video understanding
• Vals Index: first place
• Zapier's AutomationBench: first place, at 51.3%
• CWE-bench v1: tied for first at 68%, on fixing security vulnerabilities

For trusted defenders and Google's own teams, Argon will ship without cyber guardrails. Google says Wiz used the model to find a critical vulnerability in healthcare software used by hospitals worldwide, which earlier models had missed.

Google also described internal uses:

• Beating a published quantum-algorithm baseline by 40%
• Memory optimizations across its data centers that have freed more than 300 TiB so far
• Migrating C/C++ code to Rust, including the 800K+ line Fuchsia Zircon kernel. That work is still being audited.
• Rewriting the libgav1 video decoder, which now runs 2.7x faster than the earlier Rust port

These results come from Google and have not been independently checked. The source text is cut off partway through the cybersecurity section.