Alphabet On Wednesday, Gemini 4 Argon was launched, touted as the most advanced artificial intelligence model to date, bringing significant improvements in programming, network security, and complex professional tasks.

The company stated that this model has set a new record in real-world software engineering tasks, tied for first place in cybersecurity, and also led in another benchmark test that measures performance in financial, legal, and other professional tasks.

Google claims that Argon has been internally used to optimize the memory of its data centers, freeing up hundreds of TB of memory without the need to purchase additional hardware. The company also stated that quantum computing researchers have also utilized this model.

Google stated that it plans to launch this new model in phases, initially targeting trusted cybersecurity partners, and is also conducting a security assessment with the U.S. government before its release.

The day before this release, CEO Sundar Pichai ( Sundar Pichai ) met with several technology executives at the White House and then signed a voluntary security agreement with U.S. President Donald Trump to address the growing concerns regarding AI security.

Over the past year, Google has been focused on expanding faster and more cost-effective Flash models. Gemini 4 represents a major advancement for the company in cutting-edge fields, and being just over a year away from Gemini 3, this tech company is once again at the forefront of AI model competition.

Google stated on Wednesday that before the public release, the company is focusing on expanding protective measures in four key areas, including abuse and injection alerts.

Google claims that the Argon model represents a significant improvement in network security capabilities compared to the 3.8 Flash Cyber model released earlier this month, and it also performs better in vulnerability detection.

Google also stated that Argon is on par with OpenAI in terms of cybersecurity assessment benchmarks, with a score of 4.7 in the categories GPT-6, Astra, and Grok. Additionally, it leads GPT-6, Astra, and Anthropic with a score of 5.1 in Vals Index.