Google Gemini 4 Argon is now available on Omnifact via Google Cloud in the EU and Google API.
Feature Drop: Google Gemini 4 Argon Now Available
Published on October 13th, 2026
Google's first Gemini 4 model, Gemini 4 Argon, is now available on Omnifact. Released broadly today as Google's new flagship reasoning model, Argon is built for sustained, multi-step work across software engineering, financial and legal analysis, and large document portfolios1. While Gemini 3.5 Flash and Gemini 3.8 Flash remain the fast choices for everyday prompts, Gemini 4 Argon steps in when an assignment requires deep verification and long-horizon planning.
Built for Sustained Reasoning and Deep Work
Earlier Pro releases like Gemini 3.1 Pro introduced deep-thinking capabilities in preview, but complex professional tasks often hit output limits or lost precision across hundreds of pages. Gemini 4 Argon replaces those preview constraints with a production-ready sparse mixture-of-experts architecture trained specifically for multi-step problem solving1.
On Google DeepMind's October 2026 evaluations, Argon ranks first on the Vals Index at 68.9%, measuring real-world performance across finance, coding, legal, and tax tasks1. It scores 65.4% on Vals Finance Agent v2 for multi-step financial research, 51.3% (#1) on AutomationBench for end-to-end business workflow execution, and 77.9% on DeepSWE v1.1 for long-horizon software engineering1. In practice, that translates to fewer manual follow-ups when an agent searches a Knowledge Base, cross-checks regulatory filings, or refactors code across multiple files.
Gemini 4 Argon in the Google Lineup
With Gemini 4 Argon joining the lineup, organizations can pair Argon for heavy analytical work with Gemini 3.8 Flash for fast daily tasks:
| Area | Gemini 4 Argon | Gemini 3.1 Pro | Gemini 3.8 Flash |
|---|---|---|---|
| Primary Role | Deep reasoning, complex coding, legal and financial synthesis | Previous Pro reasoning baseline (preview) | Fast everyday assistance and high-volume tasks |
| Context Window | 1,048,576 tokens (unified input, thinking, and output) | 1,048,576 tokens | 1,048,576 tokens |
| Max Output | 65,536 default / up to 1,048,576 (long decode) | 65,536 tokens | 65,536 tokens |
| Reasoning Effort | Low, Medium (default), High | Configurable thinking | Configurable thinking |
| Token Pricing (per 1M) | $2.00 input / $10.00 output (intro through Nov 30, 2026) | $2.00 input / $12.00 output | $0.75 input / $3.75 output (intro through Dec 31, 2026) |
| Knowledge Cutoff | March 2026 | January 2025 | March 2026 |
Google offers Gemini 4 Argon at an introductory rate of $2.00 per million input tokens and $10.00 per million output tokens (including thinking tokens) through November 30, 20262. Starting December 1, 2026, standard list pricing of $4.00 per million input tokens and $20.00 per million output tokens applies2.
Put the 1M Context Window to Work
Gemini 4 Argon shares a 1-million-token budget across your prompt, thinking steps, and output, maintaining 84.2% recall all the way up to 1 million tokens on GraphWalks31:
- Document synthesis in Spaces: Analyze full contract suites, audit reports, and financial charts in Spaces synced via Connected Sources in a single conversation.
- Multi-file coding and security reviews: Run large-scale code migrations, dependency upgrades, and vulnerability audits across whole repositories.
- Adjustable Reasoning: Switch between Low, Medium (default), and High using the Reasoning controls in the Chat Input model picker to match thinking depth to the task3.
Available Today with EU Data Residency
Gemini 4 Argon is available today for organizations on Omnifact via Google Cloud in the EU (eu multi-region) as well as the direct Google API4. When routed through Google Cloud in the EU, all model inference takes place within European data centers under zero data retention commitments. In the Omnifact web app, every prompt, retrieved document, and thinking step is also protected by our Privacy Filter, which uses rule-based and AI-powered detection to pseudonymize personal data before it reaches any external model.
Team administrators can enable Gemini 4 Argon today in Team Settings > Models. Because Argon is a frontier reasoning model, admins can activate it for the entire workspace or assign access to specific departments using the Premium Models permission in Group Management. Once enabled, select Gemini 4 Argon in the Chat Input model picker to put it to work on your hardest assignments.
