Technology Desk: Google has introduced its new and extremely powerful AI model Gemini 4 Argon. Company CEO Sundar Pichai, while sharing the information on social media platform
According to Sundar Pichai, Google's internal teams are already using Argon extensively for complex tasks like coding and quantum computing. This is the first time Google has used 'Argon' in the name of one of its AI models.
Output limit of 10 lakh tokens
Google Gemini 4 Argon is designed to handle deep thinking and longer tasks in complex workflows. The output token limit of the model is stated to be 10 lakh tokens. Due to such a large limit, it is capable of reasoning up to hundreds of thousands of tokens in long and complex tasks.
According to the company, Argon has performed equal to or better than advanced AI models like GPT-6 Astra and Claude Fable 5.1 in many benchmarks.
Strong performance in many benchmarks
The Gemini 4 Argon achieved a score of 68.9% in the Walls Knowledge Work Benchmark. Whereas GPT-6 Astra's score is said to be 63.1% and Claude Fable 5.1's score is said to be 65.8%.
Argon also achieved a score of 91.9% in Vibe Code Benchmark related to coding. In comparison, Claude Fable 5.1 scored 90.3% and GPT-6 Astra scored 89.6%.
However, Google's new model hasn't taken the lead in every benchmark. According to company data, Argon has also lagged behind other advanced models in some other tests.
Also used in cyber security
Cyber security is also included in the key capabilities of Gemini 4 Argon. Google says that the model is capable of finding software vulnerabilities on its own, verifying them, and preparing necessary patches.
Google's cyber security company Wiz is also using Argon. During a demo the model identified a serious security weakness in healthcare software. According to Google, previous Frontier models were not able to detect this weakness.
Through this capability, Google is working towards expanding Argon beyond general AI assistant and using it in important areas like cyber defense and software security.
Not yet available to general users
Currently, the availability of Gemini 4 Argon is limited. It has been made available to selected cyber defenders and the US government through the Fairwind program.
According to Sundar Pichai, the rollout of the model is being done in a phased manner. In the future its access will be extended to paid API customers and Google AI Ultra subscribers.
What is the starting price of Gemini 4 Argon?
According to Google, the starting price for input has been kept at $2 per million tokens, which works out to approximately Rs 190 in Indian currency.
Whereas the output price is $10 per million tokens i.e. around Rs 950 per million tokens.
Prices may be increased later. In that case the input price could reach $4 per million tokens and the output price could reach $20 per million tokens.
Google's new bet in the AI race
The launch of Gemini 4 Argon isn't just the introduction of a new AI model for Google. The company is moving towards using it in areas like coding, complex professional work, cyber security and quantum computing.
At present, due to its limited availability, its widespread use has not started for common users. However, early benchmarks and cybersecurity claims released by Google suggest that the company is targeting its new model for high-end AI and complex professional workflows.