Alphabet on Wednesday unveiled Gemini 4 Argon, its most advanced artificial intelligence model yet, delivering major improvements in coding, cybersecurity and complex professional work.
The company said the model sets a new record in real-world software engineering, ranks first in cybersecurity and leads another benchmark in performance measurement in finance, legal and other business tasks.
Argon is already used internally to optimize memory in Google’s data centers, freeing up hundreds of terabytes of memory without purchasing additional hardware, the company said. Quantum computing researchers have also used this model.
Google plans to launch the new model in stages, starting with trusted cybersecurity partners while working with the U.S. government on preliminary security assessments.
“Starting this deployment in this way gives us more confidence, but also allows us to get a trained, robust cyber defense model into the hands of defenders as soon as possible,” Tulsee Doshi, product manager of Google’s Gemini model, told CNBC.
The model’s release comes a day after CEO Sundar Pichai signed a voluntary agreement with President Donald Trump and top technology executives at the White House, following a luncheon aimed at addressing growing AI safety concerns.
Gemini 4 also launches nearly a year after Gemini 3 put the tech company back at the forefront of the heated race for AI models. The long-awaited model represents a major breakthrough at the frontier and follows Google’s pivot toward scaling faster, less expensive flash models.
Google said Wednesday it plans to expand its protections in four key areas of cybersecurity, including misuse and rapid injection, before launching them publicly.
Doshi called Gemini 4 an “incredibly comprehensive” model that excels at completing long, multi-step tasks. The new model also represents a significant advancement in cybersecurity over Google’s Flash Cyber 3.8 model released earlier this month, outperforming in vulnerability discovery.
“We really believe that a model of this caliber and this level of cutting-edge performance is of significant importance to defenders,” Doshi said.
Argon is tied with GPT-6 Astra and OpenAI’s Grok 4.7 on cybersecurity benchmarks and ahead of GPT-6 Astra and recent Anthropic models on the Vals index.
Gn headline
