Gemini 4 Argon is now available on Google Cloud. The announcement positions it as a frontier model for long, multi-step work, trained for depth and precision and able to maintain context across tasks reinforced with domain expertise. Argon can read long documents mixing text, tables, charts, and images, and turn them into one grounded output. It can pull findings from many sources into a single structured answer, analyze data over time, and draft legal, finance, and executive documents. The model is also designed to respect system boundaries, report negative findings honestly, pause before irreversible steps, and tell users when it could not find something or finish a task.

The announcement groups the value into four areas. First, Argon is meant to take over deep work that previously took hours or days. Customers are reported to see strong results turning unstructured inputs into cohesive outputs; GitLab VP Stuart Moncada says it is designed to help organizations bring AI into development, security, and operations within existing enterprise workflows and controls. Second, Argon targets finance, legal, tax, and other regulated work. It leads the Vals Index for economic impact across finance, coding, legal, and tax, and can automate multi-step ledger reconciliations, spreadsheet formulas, and variance analyses. In legal research it leads Harvey’s Legal Agent Benchmark, with strong citation fidelity; Harvey’s Niko Grupen describes substantial improvements over Gemini 3.1 Pro across legal agent evaluations, especially in corporate M&A, data privacy, cybersecurity, and EC/VC. ServiceNow also reports roughly 20% improved task success over Gemini 3.8 Flash on NowAI-Bench, with standout gains in HR and customer service.

Third, Argon is positioned for faster product development. The open-source Agent Development Kit (ADK) team used Argon autonomously overnight within automated testing guardrails; the resulting optimizations, now live in google/adk-python, cut in-memory session prep time by 40% in long sessions and speed up cold imports by 21%. Google engineering teams using Argon in Google Antigravity increased agentic code submissions by more than 35% while cutting code rollbacks by nearly a third. The Chrome team used Argon to finish an architectural refactor across hundreds of files in Chromium’s 40-million-line codebase in weeks rather than months. Figma cites Argon’s handling of complex multi-step tasks in Figma Make, and AlphaEvolve uses Argon for code optimization across workloads like factory scheduling, logistics, game code speed-ups, and chip layout.

Fourth, the model is presented as a defensive security asset. Argon includes cyber mitigations that block offensive tasks such as pentesting and exploitation, but it can be used defensively through CodeMender or custom security harnesses. It ties for first place on CWE-Bench v1 with a score of 68% for remediating security vulnerabilities. HubX reports the model identified 100% of top-priority vulnerabilities with 0% fabricated vulnerabilities across verified claims in MarketingX. Argon is also described as the most resilient model yet against indirect prompt injections, leading the Gray Swan benchmark with attack success reduced to 0.7% across 15 adaptive attempts. The article also notes that Argon is being released without cyber guardrails to select trusted defenders vetted through the Fairwind Program.

Argon is available starting today in Gemini for business on the web, desktop, and mobile. Enterprise developers and builders can access it via API on the Gemini platform under model ID gemini-4-argon, in Google AI Studio, and in Google Antigravity. The introductory rate is $2 per million input tokens and $10 per million output tokens through November 30, 2026, with cached input tokens at 95% off; standard pricing afterward is $4 per million input tokens and $20 per million output tokens. The post points to the model card and developer guide for token limits, migration instructions, and technical specifications, and invites security professionals to the Security Summit for more on using Argon to defend against machine-speed threats and automate remediation.