Google has made its Gemini 4 Argon model available on Google Cloud, extending access across Gemini business and developer tools.
The launch brings Argon to Gemini for business on web, desktop and mobile. Developers can also use it through an API in the Gemini platform, Google AI Studio and Google Antigravity.
Google positioned Argon for long, multi-step work in areas including software development, legal research, finance and cyber defence. The model can process long documents containing text, tables, charts and images, and produce outputs tied to those source materials.
Google is also introducing a two-tier pricing structure. Argon starts at USD $2 per million input tokens and USD $10 per million output tokens at an introductory rate, before moving to USD $4 per million input tokens and USD $20 per million output tokens under standard pricing.
Customer Testing
Several partners and customers cited results from early access and evaluation work. Google said users had tested the model in software development, legal analysis, human resources, customer service and security review.
“As software development becomes more agentic, teams need models that can support long-running, multi-step work across all stages of the development lifecycle. Gemini 4 Argon is designed to help organizations bring AI into development, security and operations while keeping it within established enterprise workflows and controls,” said Stuart Moncada, Vice President, GitLab.
Harvey, which focuses on legal work, said the latest model outperformed an earlier Google model in its internal assessments.
“Gemini 4 Argon shows how much progress Google has made on agentic legal reasoning. We saw substantial improvements over Gemini 3.1 Pro across our legal agent evaluations, particularly in corporate M&A, data privacy and cybersecurity, and EC/VC. Accuracy also held up on long-document analysis, an encouraging result for lawyers working through detailed transactional matters,” said Niko Grupen, Head of Applied Research, Harvey.
ServiceNow said the model delivered stronger results in selected enterprise tasks during early access.
“We collaborated with Google on NowAI-Bench to help shape Gemini 4 for key enterprise work. In early access, it delivered roughly 20% higher task success than Gemini 3.8 Flash, with standout gains in HR and customer service. On the ServiceNow AI Platform, our L1 Service Desk AI specialist brings Google's most advanced intelligence, governed and secure, turning tasks into fully autonomous work,” said Bardoliwalla.
Internal Use
Google also disclosed internal engineering uses for Argon. The team behind its open-source Agent Development Kit used the model to identify optimisations that reduced in-memory session preparation time by 40% in long sessions and improved cold imports by 21%.
Across Google, engineering teams using Argon in Google Antigravity increased agentic code submissions by more than 35% while reducing code rollbacks by nearly a third, according to the company. Google also said the Chrome team used the model to complete an architectural refactor across hundreds of files in the Chromium codebase.
Figma described the model as useful for more complex product design work.
“Gemini 4 Argon handles complex tasks reliably, from 3D games to rich animations. In Figma Make, it helps teams turn their most ambitious, multi-step ideas into impressive prototypes,” said Loredana Crisan, Chief Design Officer, Figma.
Security Focus
Cybersecurity is another part of the rollout. Google said Argon includes restrictions that block offensive cyber tasks such as pentesting and exploitation, while still allowing defensive work through CodeMender or other security tools.
Google said the model tied for first place on CWE-Bench v1 with a score of 68% for remediating security vulnerabilities. It also led the Gray Swan Indirect Prompt Injection benchmark, with attack success reduced to 0.7% across 15 adaptive attempts.
HubX said its testing focused on codebase audits.
“We are integrating Gemini 4 Argon into MarketingX for automated codebase security audits. In our evaluations, the model successfully identified 100% of top-priority vulnerabilities with 0% fabricated vulnerabilities across verified claims. This precision unlocks proactive, code-level security for our customers before deployment,” said Yunus Emre, Chief AI Officer, AI Lab, HubX.
Google said a version without cyber guardrails will be released to a select group of vetted defenders through its Fairwind Program, which serves government, critical infrastructure and cybersecurity partners.