Google has unveiled Gemini 4 Argon, a new AI model that the company says is already being used extensively by its own engineers, even though outside users cannot access it yet. The release is limited, and Google has not announced API pricing because the model remains in testing. The company says it is taking a phased approach, starting with a small group of trusted testers.
According to Google, Argon has delivered internal results. The company says the model used fleet-wide telemetry data to help save 300 TiB of memory across its data centers. Google also says Argon agents have migrated C/C++ codebases to Rust, including thousands of lines in the core re2 and libgav1 libraries and more than 800,000 lines in the Fuchsia OS Zircon kernel.
Google brought benchmarks to support its claims. On the software engineering DeepSWE v1.1 benchmark, Gemini 4 Argon scores 77.9 percent, which Google says is higher than GPT-6 Astra, Fable 5.1, and Opus 5.5. The company also points to an industry-leading score on the Vals Index, an economic analysis test, and promises similar capability across long-horizon tasks.
The confirmed output limit for Gemini 4 Argon is 1 million tokens, up from 64,000 tokens in previous Gemini models. Google says the higher ceiling lets users complete more demanding tasks in a single step.
Cyberdefense is the main focus for Gemini 4 at this stage. Google says the size of the model requires a phased release, and partners in its Fairwind Program can get access to use the model's cybersecurity defense capabilities. Wiz is already using Argon, according to Google, and used it to find a critical vulnerability that could expose personal information in a system used at hospitals around the world. Google claims other frontier models missed the flaw but did not provide specifics.
The limited release comes after a summer of high-profile hacking incidents that have raised concern about model misalignment. Google says Argon was designed with systems that monitor the model's chain-of-thought and can halt it if it steps out of bounds. The company says that capability is why reasoning transparency matters in frontier model development.
For US technology watchers, the announcement signals that access to Google's most advanced model will remain restricted for now, with pricing and broader availability still undecided.
More AI news from TechManNews.
