Google Unveils Gemini 4 Argon With One Million Token Output Limit
Google has introduced Gemini 4 Argon, a flagship artificial intelligence model engineered for complex, multi-step workflows spanning software engineering, financial research, legal analysis, and cybersecurity. The model anchors Google's new Gemini 4 generation, replacing previously planned interim releases such as Gemini 3.5 Pro. Google positions Argon to compete directly with frontier systems from OpenAI and Anthropic by emphasizing extended reasoning capabilities, cost competitiveness, and deep cloud ecosystem integration.
A foundational advancement in Gemini 4 Argon is its expanded output capacity, which reaches up to one million tokens in a single trajectory compared to 64,000 tokens in previous versions. According to Google, this large output window enables the model to sustain complex problem-solving chains over long-running workflows without requiring frequent session splits or premature task termination. For context, Anthropic's Claude Opus 5.5 supports up to 300,000 output tokens, while OpenAI's GPT models support up to 128,000 maximum tokens per request.
Enterprise Applications and Benchmarking Performance
Google reports that Argon excels across professional evaluation metrics, leading the Vals Index for multi-domain tasks covering finance, coding, legal work, and tax analysis. The model achieved a score of 77.9 percent on DeepSWE v1.1 for long-horizon software engineering, 51.3 percent on AutomationBench for end-to-end business execution, and 91.7 percent on LVBench for long-video comprehension. Internal deployment at Google includes quantum research optimization, code migration from C and C++ to Rust, and infrastructure memory management.
Commercial availability will follow a staged rollout. Google is initially releasing Argon to trusted cyber defenders and testers through the Fairwind program and participating in U.S. government pre-release evaluations. Subsequent general availability will target paid API customers and Google AI Ultra subscribers. Introductory API pricing is structured at $2 per million input tokens and $10 per million output tokens, with subsequent rates scheduled to double to $4 and $20 per million tokens respectively.
According to SecurityWeek, google announced the new frontier model, Gemini 4 Argon.
Cybersecurity Capabilities and Monitored Autonomous Execution
A central focus of the Argon release is automated vulnerability remediation. Google has trained the model to autonomously identify, validate, and patch critical software flaws, offering unguardrailed versions to selected defensive partners such as Wiz under its Scan for Good initiative. The model achieved a score of 68 percent on CWE-bench v1 for security vulnerability remediation and successfully identified a critical healthcare software vulnerability missed by earlier models.
To manage autonomy risks, Google has deployed supervisory monitoring systems designed to track reasoning trajectories and halt execution when model actions deviate from user intent. Testing protocols also feature sandboxed isolation and enhanced resilience against indirect prompt injection attacks as the company prepares broader commercial deployment.