
What happened
Google announced Gemini 4 Argon on October 1, priced at $2 per 1M input tokens and $10 per 1M output tokens, with cached input tokens at a 95% discount. It rolls out first to trusted cybersecurity professionals via the Fairwind Program.
Why it matters
The 1M output limit means users can generate hundreds of thousands of tokens in one process, supporting deeper reasoning. Coding, knowledge work, cybersecurity, and creative writing are the targeted frontier-level uses.
What to watch
Whether the cybersecurity-focused rollout limits broader enterprise access remains unstated. The internal vulnerability benchmark found flaws across 20 programming languages, so real-world abuse prevention is the test.
WHO IT HITSCybersecurity professionals gain early access through the Fairwind Program, while enterprise developers and professional users evaluating frontier models will weigh the $2 per 1M input token and $10 per 1M output token pricing against competitors.
Summaries like this, every morning.
Google's announcement of Gemini 4 Argon on October 1 positions the model as a frontier-level tool for coding, knowledge work, cybersecurity, and creative writing. The company is not releasing it broadly at once; instead, it is using the Fairwind Program to give trusted cybersecurity professionals sequential access, a controlled approach that echoes how sensitive capabilities are sometimes staged. Pricing is set at $2 per 1M input tokens and $10 per 1M output tokens, with cached input tokens at a 95% discount, which may appeal to heavy users who reuse prompts.
The model's expanded output limit—from 64,000 tokens to 1 million tokens—is a technical shift that Google says enables generating hundreds of thousands of tokens in one process while reasoning deeply. Google also reports progress in migrating C/C++ codebases to Rust: for the video decoder libgav1, Gemini 4 Argon generated code 2.7 times faster than an existing Rust port. On benchmarks, it scored 77.9% on DeepSWE v1.1 for software development and 91.7% on LVBench for long-video understanding.
On cybersecurity, Google says its internal comprehensive vulnerability benchmark found a wide range of vulnerabilities across complex codebases spanning 20 programming languages. The model is also designed to refuse harmful requests to prevent misuse for cyberattacks or chemical, biological, radiological, and nuclear attacks. Whether this controlled rollout and safety design are enough to satisfy enterprise buyers evaluating the model's real-world risk remains to be seen.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
IMPACTLAKE began offering a ChatGPT plugin for impactlake(TM) and added Model Context Protocol (MCP) support…

Ace Pillar, a Qisda Group affiliate, reported strong third-quarter 2026 results as its semiconductor, AI autom…

Microsoft entered a new AI partnership with Cerebras, whose AI hardware and software will be used within Micro…

The author installed the open-source agent platform OpenClaw on an EPSON Endeavor NY2100S (Celeron P4600, 4GB…

An analysis lays out how Dell Automation Platform uses a portal, catalog and orchestrator to apply reusable bl…

A developer using Claude Code on a personal web app discovered via --debug logs that his Stop hook passed ever…
