Google Unveils Gemini 4 Argon AI Model, Unavailable to Public

Google has introduced Gemini 4 Argon, its latest artificial intelligence model, designed to enhance performance in coding, knowledge work, and cybersecurity. The company stated that while the model is not yet available for public use, Google engineers are extensively testing it internally.

Argon is reported to have already contributed to operational efficiencies within Google's data centers, saving an estimated 300 terabytes of memory by utilizing fleet-wide telemetry data. The AI has also been involved in migrating Google's C/C++ codebases to the Rust programming language, including substantial contributions to core libraries like re2 and libgav1, as well as the Fuchsia OS Zircon kernel.

Google presented benchmark results to support its claims of Argon's capabilities. The model reportedly achieved a 77.9 percent score on the DeepSWE v1.1 software engineering benchmark, surpassing other models such as GPT-6 Astra, Fable 5.1, and Opus 5.5. The company also highlighted Argon's performance on the Vals Index test for economic analysis, suggesting broad applicability in long-horizon tasks.

Details regarding the public release timeline for Gemini 4 Argon have not yet been disclosed by Google. The company previously announced Gemini 3.5 Pro in June and has since introduced smaller Flash models.