Google Teases Gemini 4 Argon: New AI Model Outperforms ChatGPT-6 and Claude Opus 5.5
- Google is rolling out Gemini 4 Argon, a frontier AI model that delivers sustained deep reasoning across complex, long-horizon workflows and outperforms competing systems from Anthropic and OpenAI...
- Google reports that Gemini 4 Argon achieves frontier-level performance in complex real-world tasks, outperforming rival systems such as Claude Opus 5.5, Fable 5.1, and ChatGPT-6 Astra.
- Google is actively utilizing Gemini 4 Argon within its internal workflows, noting that the model is fundamentally changing daily operations.
Google is rolling out Gemini 4 Argon, a frontier AI model that delivers sustained deep reasoning across complex, long-horizon workflows and outperforms competing systems from Anthropic and OpenAI across multiple benchmarks.
Gemini 4 Argon Performance and Benchmarks
Google reports that Gemini 4 Argon achieves frontier-level performance in complex real-world tasks, outperforming rival systems such as Claude Opus 5.5, Fable 5.1, and ChatGPT-6 Astra. The model also leads the Vals Index, which measures AI models for their economic value across a wide range of sectors. Google states that Argon is designed to maintain deep reasoning over extended workflows, marking a significant performance shift from previous iterations.
Internal Deployments and Operational Efficiency
Google is actively utilizing Gemini 4 Argon within its internal workflows, noting that the model is fundamentally changing daily operations. A team of Argon agents applied memory optimizations across Google data centers, freeing up 300 TiB of storage space, with total estimated savings reaching between 500 TiB and 1 PiB. Argon agents are also assisting in migrating C and C++ databases to Rush by working through tens of thousands of lines of core library code. Quantum computing researchers at Google also used Argon to optimize spacetime resources, with one agent beating the published baseline by 40% in minutes.
Token Limits and Pricing Structure
Alongside the model release, Google is introducing support for a 1-million output token limit, a substantial increase from the 64K token limit of previous Gemini versions. This expanded capacity allows the model to process longer and more complex tasks and generate extensive outputs in a single response. For API users, Google is offering introductory pricing of $2 per million input tokens and $10 per million output tokens, alongside a 95% discount on cached tokens. After the introductory period concludes, standard pricing will double to $4 per million input tokens and $20 per million output tokens.
Rollout Plan and Access Requirements
Gemini 4 Argon is initially rolling out to trusted cyber defenders through the Fairwind Program as Google works to strengthen model safeguards prior to a broader public release. Following feedback and safety refinements from testers, access will expand to paid API customers and Google AI Ultra subscribers.