Google Launches Gemini 4 Argon With 1M Token Output and Low Prices

Google has launched Gemini 4 Argon, marking its return to the frontier AI race after a period of trailing rivals. The company claims the new model delivers improved performance across complex workflows while reducing common issues such as repetition and hedging in generated text.

According to The Verge, Google states that Argon competes with industry leaders like GPT-6 Astra and Claude Opus 5.5. The firm reports superior results in benchmarks covering coding, cybersecurity, science, finance, and law. This release follows an initial limited rollout to trusted cyber defenders for risk testing.

Technical Specifications and Performance

MarkTechPost reports that Gemini 4 Argon is a sparse mixture-of-experts transformer trained on TPUs using JAX and ML Pathways. It features native multimodal support for text, images, audio, and video input, though it outputs only text. Google has not disclosed total or active parameter counts.

The model leads Google’s internal tables on knowledge work, long context, and video understanding. It achieves 91.7 percent on LVBench for long video understanding and tops the Vals Index at 68.90 percent. However, it scored last among four models on FrontierSWE v2 with 55.0 percent, indicating mixed results in agentic coding tasks.

Availability and Pricing Structure

The model is available today to Pro and Ultra subscribers in the Gemini app, AI mode in Search, and Ultra subscribers on the Antigravity platform. It is also accessible via the Gemini API and AI Studio for developers and enterprises. Google has not confirmed plans to release Argon to Plus subscribers or free users.

Introductory pricing is set at $2 per million input tokens and $10 per million output tokens. This is approximately one-fifth the cost of competing models from OpenAI and Anthropic. Prices will double at the end of November, though Argon will remain cheaper per token than rivals. A full one-million-token response costs $10 at introductory rates.

What should you do?

Security and technology professionals should evaluate whether Argon’s long-context capabilities suit their specific workflow requirements, particularly for document-heavy tasks. Developers should monitor the pricing changes scheduled for late November to adjust budget forecasts accordingly. Teams relying on coding assistance should review the mixed performance data on software engineering benchmarks before full integration.

What to do and how to stay safe: Google

• Verify that your organization’s data privacy policies allow sending sensitive information to cloud-based models without open weights.
• Test the model’s output length limits by running pilot projects with documents approaching the one-million-token window.
• Review the model card for the March 2026 knowledge cutoff to ensure recent data requirements are met.
• Monitor official channels for updates on availability for lower-tier subscription plans.

Step-by-step guide: Mobile Device Security: How to Protect Data Beyond Passwords

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Is Gemini 4 Argon available to free users?

No, the model is currently available only to Pro, Ultra, and API subscribers, with no confirmed plans for free users.

What is the maximum output length for this model?

Gemini 4 Argon can generate up to one million tokens in a single response, shared within a combined input-output window.

How does the pricing compare to competitors?

Introductory pricing is $2 per million input tokens and $10 per million output tokens, roughly one-fifth the cost of rival models.