AI news
Google DeepMind Unveils Gemini 4 Argon with 1--Million Output Token Limit
Date: September 30, 2026
Sources: Google Blog, NeuralTrust, WorkOS
Strategic Overview
Google DeepMind has introduced Gemini 4 Argon, the flagship opening model of its next--generation Gemini 4 frontier architecture. Engineered for complex, multi--step agentic execution and autonomous enterprise workflows, Argon features an unprecedented single--response output cap of 1 million tokens—a massive leap from the standard 64,000--token boundaries of previous iterations.
Key Facts & Enterprise Impact
- Extended Reasoning & Output Capacity: The 1--million--token output limit allows enterprise agents to conduct exhaustive test--time compute, execute full--codebase refactoring, and produce long--horizon analytical audits in a single continuous stream.
- Performance Benchmarks: Google reports a 77.9% score on software engineering evaluation DeepSWE v1.1, positioned alongside Claude Opus 5.5 and GPT--6 Astra in enterprise knowledge work and complex code synthesis.
- Phased Deployment via Fairwind: Addressing dual--use safety risks, Google has restricted initial access to vetted cybersecurity specialists through its Fairwind Program, providing defensive teams with ungated capabilities under strict identity controls while enterprise API and consumer tiers remain queued.
- Commercial Pricing: Google launched Argon with an introductory pricing schedule of $2.00 per million input tokens and $10.00 per million output tokens, alongside a 95% discount on prompt--cached tokens.