On 30 September 2026, Google announced its new flagship model, Gemini 4 Argon. The official post frames Argon as a “next era of frontier intelligence,” focused on complex, long-horizon workflows in real-world software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense. Rather than a same-day consumer free-for-all, Google describes a phased path: Argon is rolling out first to a set of trusted cyber defenders through the Fairwind Program, while the company gathers early tester feedback and strengthens guardrails before making the model available to developers, enterprises, and consumers as soon as possible—starting with paid API customers and Google AI Ultra subscribers.
Why cyber defenders first
Google says safely releasing frontier capabilities at this level requires a phased approach. Argon is rolling out to trusted cyber defenders via Fairwind, and the company states it is engaged in the U.S. government’s voluntary process for pre-release model access while gradually expanding access.
On defensive cybersecurity, Google says Argon can autonomously find, validate, and patch critical software vulnerabilities. For trusted defenders and Google’s own internal teams, Argon is to be released without cyber guardrails so they can use its full frontier-level defensive capabilities. Cloud security firm Wiz is already using Argon through its Scan for Good initiative; Google cites an early demonstration in which the model uncovered a critical vulnerability exposing sensitive personal information in healthcare software used by hospitals worldwide—a severe risk Google says previous frontier models had missed. On CWE-bench v1, which evaluates remediating security vulnerabilities, Google says Argon ties for first place at 68%.
Long-horizon work: up to 1M output tokens
To support longer, more complex use cases, Google is expanding Argon’s output token limit to what it calls an industry-leading 1 million tokens, up from the previous 64K. The stated rationale is that when the model has headroom to think deeply and generate hundreds of thousands of tokens in a single trajectory, it can solve tougher problems in one go.
On coding and enterprise workflows, Google engineers have used Argon from everyday debugging to large-scale codebase migrations and algorithm design. Company-reported results on the official page include DeepSWE v1.1 at 77.9% for real-world long-horizon software engineering; a leading position on the Vals Index (economic impact across finance, coding, legal, and tax work); AutomationBench (Zapier’s end-to-end business-function benchmark) at 51.3%, ranked #1; and LVBench for long-video understanding at 91.7%. These figures are Google’s published claims; readers should treat them as company-reported and consult the original post and benchmark owners for methodology.
Internal examples Google cites
The blog lists Argon already powering internal Google workflows, including:
- Quantum algorithmic optimization: helping optimize spacetime resources (qubits × gates) of bottlenecking subroutines; in one example, beating a published baseline by about 40% in minutes.
- Memory efficiency: a team of Argon agents analyzed fleet-wide profiling telemetry to identify and apply memory optimizations across Google’s data centers, freeing over 300 TiB once rolled out, with an estimated 500 TiB to 1 PiB in total savings.
- Large-scale codebase migrations: Argon agents working on migrating C/C++ codebases to Rust—from tens of thousands of lines in core libraries up to 800K+ lines for the Fuchsia Zircon kernel—with rigorous automated and manual auditing before production. For libgav1, Google’s open-source video decoder, agents replaced about 32K lines of SIMD code through profile-guided experiments, producing safe Rust that the compiler vectorizes automatically; Google reports a memory-safe decoder about 2.7× faster than the prior Rust port, with identical video output.
Four safeguard axes before broad availability
Before broader rollout, Google says it is strengthening frontier safeguards across four areas:
- Defending against misuse: under the Frontier Safety Framework, the model is designed to refuse harmful requests (including cyber or chemical, biological, radiological, and nuclear attacks) while preserving legitimate dual-use scientific research; Google is improving techniques such as monitoring internal activations, with robustness testing by internal and external red teams.
- Defending against prompt injection: Google calls Argon its most resilient model yet against indirect prompt injections, and reports a leading position on Gray Swan’s Indirect Prompt Injection (IPI) benchmark after automated red teaming and adversarial training.
- Monitoring for misalignment: deploying mitigations that monitor Argon’s chain-of-thought and actions and stop execution when necessary; similar monitoring during training alerts an incident-response team, with care not to feed findings back into training in ways that could teach the model to evade monitors. Google encourages the industry to preserve reasoning transparency as capabilities rise.
- Hardening systems: isolating and sealing sandboxed environments before high-risk training or evaluations, in line with Google’s agent control roadmap, and sharing agent-security practices with partners.
Pricing and who can use it
Google lists an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off the input token price. After the introductory period, $4 per 1M input and $20 per 1M output will apply. Wider availability remains “as soon as possible,” starting with paid API customers and Google AI Ultra subscribers; the current front door is Fairwind trusted cyber defenders plus internal and early testers. Actual access depends on Google accounts, API status, and the official product pages.
This brief follows Google’s 30 September 2026 introduction on phased release, long-horizon output limits, and defensive cybersecurity capabilities. Benchmark scores and internal case studies are company-reported and were not independently re-run. Availability in any given plan or market should be checked against Google’s live product status.
Primary sources
- Google: Introducing Gemini 4 / Gemini 4 Argon (2026-09-30)
