Google releases Gemini 4 Argon to Pro subscribers, offering million-token outputs at introductory rates that undercut rival AI models.
Key points
- Gemini 4 Argon launches for Pro and Ultra subscribers with an introductory price of $2 per million input tokens.
- The model supports outputs up to one million tokens, a significant increase from previous Gemini versions.
- Google claims Argon outperforms GPT-6 Astra and Claude models on benchmarks in coding, law, and cybersecurity.
Google has launched Gemini 4 Argon, its newest frontier artificial intelligence model, making it available to higher-tier subscribers and developers. The release marks a return to the forefront of the AI race after months of trailing competitors.
What happened
Google announced the general availability of Gemini 4 Argon, according to The Verge. The model is now accessible to Pro and Ultra subscribers in the Gemini app and AI mode in Search. It is also available to Ultra subscribers on the Antigravity development platform. The Verge reports that availability for Plus subscribers or free users remains unclear.
MarkTechPost notes that the model is generally available on the Gemini API and AI Studio for developers and enterprises. Google positioned Argon as a sparse mixture-of-experts model designed for long, multi-step workflows. The company states it is already using Argon internally to find cybersecurity vulnerabilities and debug drone physics simulators for Wing.
Pricing for the model starts at $2 per million input tokens and $10 per million output tokens, according to The Verge. These introductory rates are approximately one-fifth the cost of OpenAI’s GPT-6 Astra and Anthropic’s Fable 5 models. Costs will double at the end of November, but the model will remain cheaper than rivals on a per-token basis.
Google claims Argon leads the Vals Index at 68.90%, according to MarkTechPost. The company also reports top scores on DeepSWE v1.1, AutomationBench, and Harvey’s Legal Agent Benchmark. However, Argon scored 55.0% on FrontierSWE v2, placing last among the four models in Google’s comparison table.
Why it matters
New frontier models often shift industry standards for performance and cost. Organizations should evaluate whether the increased output capacity aligns with their specific workflow requirements. Lower token pricing can reduce operational expenses for large-scale deployments, but actual savings depend on usage patterns. Teams typically compare benchmark results against their own internal testing data to verify claims. It is important to review the specific domains where a model excels or struggles.
The fix
- Verify subscription tier requirements before attempting to access the new model features.
- Review the updated pricing structure to anticipate costs after the introductory period ends.
- Check developer documentation for the new continuation protocol when handling long outputs.
- Assess benchmark results relevant to your specific industry before adopting the model for production tasks.
Why it matters
When evaluating new AI releases, focus on the practical changes in capability and access. Users should determine if the extended context window solves specific problems in their workflow. Pricing structures often change after launch, so planning for future costs is essential. Benchmark scores provide a comparative baseline, but real-world performance can vary based on use case. Organizations should test new models in controlled environments before full integration.
What to do and how to stay safe: Google
- Review your current AI subscription to see if you qualify for early access to new models.
- Analyze the cost difference between the introductory rate and the projected standard rate.
- Test the model on a small set of tasks to verify performance claims for your specific needs.
- Monitor official announcements for updates on availability for free or lower-tier users.
Step-by-step guide: What is password policies and how they protect systems
General security guidance from the Crypt Weekly newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Who can use Gemini 4 Argon right now?
Pro and Ultra subscribers can use it in the Gemini app and Search, while Ultra subscribers can use it on Antigravity, according to The Verge.
What is the introductory price for Gemini 4 Argon?
The model costs $2 per million input tokens and $10 per million output tokens during the introductory period, according to The Verge.
How does Argon perform compared to rivals?
Google claims it outperforms GPT-6 Astra and Claude models on most benchmarks, including coding and law, according to MarkTechPost.
