Google DeepMind announced Gemini 4 Argon, a new frontier model designed to sustain deep reasoning across complex, long-horizon workflows. The company says the model is initially rolling out to a set of trusted cyber defenders through its Fairwind Program, with broader availability planned after further testing.
According to Google, Argon delivers frontier performance in real-world software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense. The model expands the output token limit to an industry-leading 1 million tokens, up from 64K, which Google says allows deeper reasoning on difficult problems in a single trajectory.
Internally, Google reports that Argon is already powering workflows across thousands of employees. Examples cited include quantum algorithmic optimization, where the model beat a published baseline by 40% in minutes; memory optimization across data centers that freed over 300 TiB of memory with estimated total savings of 500 TiB to 1 PiB; and large-scale C/C++ to Rust migrations, including the Fuchsia OS Zircon kernel at over 800K lines. For the libgav1 video decoder, Argon agents replaced 32K lines of SIMD code, producing a memory-safe decoder that runs 2.7x faster than the prior Rust port with identical video output.
On benchmarks, Google says Argon sets a new state of the art on DeepSWE v1.1 at 77.9%, leads the Vals Index, and ranks first on Zapier's AutomationBench with 51.3%. It also reports state-of-the-art long video understanding on LVBench at 91.7%. In cybersecurity, Argon ties for first on CWE-bench v1 with 68%, and Wiz is using it through its Scan for Good initiative, where it reportedly uncovered a critical vulnerability in healthcare software that previous frontier models missed.
Google says Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off the input token price. The company is engaged in the U.S. government's voluntary pre-release model access process and is prioritizing safety and rigorous testing before a wider release.
Before broad availability, Google says it is strengthening safeguards in four areas: defending against misuse for cyber or CBRN attacks, defending against indirect prompt injection, monitoring for misalignment, and preserving reasoning transparency. The model is being released without cyber guardrails to trusted defenders and internal teams to maximize defensive cybersecurity capabilities.