Google Launches Gemini 4 Argon to Rival GPT-6 Astra

Google has specifically designed this AI model for long, comprehensive, and complex tasks. The platform is currently available to trusted testers through the Fairwind program. The next in line are the API customers and Google AI Ultra subscribers who can access Google’s latest AI model .

Gemini 4 Argon is Google's AI model that will directly compete with OpenAI's GPT -6 Astra . According to the company, the model is equipped to handle complex tasks such as software engineering, enterprise work, research, and cybersecurity. Google will roll out its new model in multiple phases, starting with cyber defenders, using early feedback to improve the model's safety measures.

Today we’re introducing Gemini 4 Argon.

About Argon

Google claims Gemini 4 Argon beats GPT 6 Astra on several benchmarks. The company built the model for long tasks that need steady reasoning. Thousands of Google employees already tested it inside the company.

The model can write up to 1 million tokens in one response. The previous generation topped out at 64,000. This helps Argon handle longer jobs in a single session.

Google shared a few scores: Argon got 77.9 percent on DeepSWE v1.1, a test for real-world coding work. It also ranks first on AutomationBench with 51.3 percent. Google pointed to its quantum computing team as one example.

The model helped researchers cut the resources used by key code routines.

"It beat the published baseline by 40% in a matter of minutes," the company said.

Argon is also being used for large code migrations. Also read: ChatGPT India’s ‘Astra Uthao, Parth’ Bio Sparks Cultural Localization Buzz Rollout, Safety & Pricing Google is releasing Argon in phases. For now, only selected cyber defenders and trusted testers can use it.

Their feedback will help the company improve safety before a wider launch. Security is a big focus. Argon can reportedly find, verify, and fix software flaws.

Google also calls it its "most resilient model yet" against indirect prompt injections. These are attacks that hide harmful instructions to hijack an AI model. The company is also part of the US government's voluntary process that gives officials early access to top AI models.

Paid API customers and Google AI Ultra subscribers will get Argon first. Other developers, businesses, and everyday users will follow. Pricing starts at USD 2 per million input tokens and USD 10 per million output tokens.

Cached input tokens cost 95 percent less than the standard input price.