The architecture enables persistent problem-solving across software engineering, enterprise operations, and cybersecurity by dramatically expanding single-trajectory output generation. In internal trials, the system demonstrated massive infrastructure gains, including rewriting low-level code into memory-safe Rust and clearing over 300 TiB of data center memory. It has also captured state-of-the-art results on industry benchmarks evaluating code generation, workflow automation, and financial analysis.
Key points
The system expands output limits from 64,000 to 1 million tokens to sustain deep reasonin…: The system expands output limits from 64,000 to 1 million tokens to sustain deep reasoning during prolonged multi-step operations.
Broad availability will be priced at $2 per million input tokens with a 95% discount for…: Broad availability will be priced at $2 per million input tokens with a 95% discount for cached data, alongside $10 per million output tokens.
Specialized enterprise evaluations showed the model securing first place on Zapier’s Auto…: Specialized enterprise evaluations showed the model securing first place on Zapier’s AutomationBench as well as leading rankings on the Vals Index.
Sources checked by our analysis pipeline: 1
Gemini 4 Argon: our next era of frontier intelligence
Today, we’re announcing our new frontier model, Gemini 4 Argon, which is rolling out to a set of trusted cyber defenders through our Fairwind Program. Built to sustain deep reasoning across complex, long-horizon workflows, Argon is fundamentally changing the way we work and build at Google. It delivers frontier performance to help in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense.
Safely releasing frontier capabilities at this level requires a phased approach. We are actively engaged in the U.S. government’s voluntary process for pre-release model access while we gradually expand access. We’ll continue to gather feedback from early testers as we iterate on guardrails before making Argon available to developers, enterprises, and consumers as soon as possible.
Argon will launch at an introductory price 1 of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price.
Financial advisers preparing the artificial intelligence firm for its public debut are evaluating massive financial targets amid strong expectations in the sector. This prospective baseline demonstrates the immense scale underwriters believe the market could absorb for the enterprise. A final decision will test broader investor appetite as the listing approaches.
Developers can now run and evaluate complex cloud environments directly on their local machines without relying on remote connections. The utility was created without human intervention, marking an unprecedented milestone where automated systems independently delivered a customer-facing product. This approach highlights how major cloud providers are relying entirely on automated workflows to build developer tooling.
Designed as a model-agnostic optimization tool under the Apache 2.0 license, this release targets efficient cross-modal execution spanning text, audio, and visual generation. The system significantly accelerates quantization workflows while curbing memory overhead, enabling high-performance local serving across consumer frameworks and specialized hardware. By integrating directly into Ascend environments, it offers enterprise infrastructure substantial speedups for low-bit deployment without sacrificing model prec…