Google has unveiled Gemini 4

What's happened

Google has unveiled Gemini 4 Argon, a frontier AI model the company says leads benchmarks in coding, cybersecurity and long-horizon professional work. The model has been rolled out to trusted cybersecurity partners and internal teams while Google conducts further safety testing and readies paid subscribers for phased access.

What's behind the headline?

What this release actually does

• Gemini 4 Argon is positioned as a purpose-built frontier model for complex technical workflows: coding, large-scale engineering migrations and cyberdefense. Google is presenting benchmark wins to justify a phased, safety-first rollout.

Why Google is starting with cyber partners

• Google is placing Argon with vetted cybersecurity partners and internal engineers. That strategy will let the company gather operational telemetry and limit exposure while Argon is handling high-risk tasks like vulnerability discovery and automated patch suggestions.

The limits of benchmark claims

• Benchmarks will show Argon outperforming rivals on several indexes, but those results will not guarantee consistent real-world outcomes. Multiple outlets report internal users finding gaps when applying the model to day-to-day engineering tasks. Models frequently optimise for test suites; production work will still reveal edge cases.

Immediate consequences

• This will put powerful defensive tooling into the hands of a small set of defenders first, speeding vulnerability discovery for those partners. It will also concentrate early risk — flaws or misuses that surface in limited deployments will shape regulatory and industry responses.

What happens next

• Google will expand access after further testing, prioritising paid subscribers and enterprise customers. This phased approach will drive rapid feedback cycles inside organisations that adopt Argon early and will force competitors to respond with their own guarded rollouts or improved guardrails.

Bottom line

• Argon will accelerate defensive automation inside selected partners and Google itself, but real-world adoption will expose gaps that benchmarks do not capture. That exposure will determine whether Argon becomes a safe, reliable tool for widespread engineering and security work or a model that requires heavier operational controls.

How we got here

Google has been developing successive Gemini models after Gemini 3. The company paused a June rollout of an interim frontier model and focused on smaller Flash variants. Rising concern about AI safety and recent incidents involving agents have pushed Google to phase this release.

Our analysis

Google’s own statements appear across several pieces. TechCrunch reported Google saying Argon "can autonomously find, validate, and patch critical software vulnerabilities" and that engineers are "using the model for their daily work," quoting Google's blog post. CNBC quoted Tulsee Doshi, Google’s Gemini product lead, saying the phased rollout “gives us more confidence, but also enables us to put a model that is trained and strong in cyber defense in the hands of defenders as soon as possible.” Ars Technica cited Google’s benchmarks — for example, a 77.9% score on DeepSWE v1.1 — and noted Argon’s higher 1,000,000-token output limit and internal uses such as migrating C/C++ codebases to Rust and freeing “300 TiB” of memory. Axios placed the launch in context, saying Google has "been rolling out smaller, cheaper Flash models" and noting Bloomberg-reported internal doubts about consistent performance; Axios quoted Tulsee Doshi: "Argon is a well-rounded model that has frontier capabilities across several domains." The Japan Times and Bloomberg-sourced reporting both warned that internal users have found weaknesses when applying the model to practical coding tasks. Together, these sources show a consistent corporate message about benchmark leadership and internal utility, while independent reporting stresses that benchmarks do not fully predict real-world reliability and that some insiders see performance gaps.

Go deeper

• How will Google verify Argon’s safety before broader release?
• Which cybersecurity partners have live access to Argon?
• When will paid subscribers get API access to Gemini 4?

More on these topics

• Google - Technology company

  Google LLC is an American multinational technology company that specializes in Internet-related services and products, which include online advertising technologies, a search engine, cloud computing, software, and hardware.

• OpenAI - Artificial intelligence company

  OpenAI is an artificial intelligence research laboratory consisting of the for-profit corporation OpenAI LP and its parent company, the non-profit OpenAI Inc.

• Anthropic - Artificial intelligence company

  Anthropic PBC is a U.S.-based artificial intelligence startup public-benefit company, founded in 2021. It researches and develops AI to "study their safety properties at the technological frontier" and use this research to deploy safe, reliable models for

• Sundar Pichai - Chief Executive Officer of Alphabet

  Pichai Sundararajan, known as Sundar Pichai, is an Indian-American business executive. He is the chief executive officer of Alphabet Inc. and its subsidiary Google LLC.

• The United States of America - Country in North America

  The United States of America, commonly known as the United States or America, is a country mostly located in central North America, between Canada and Mexico.

• Alphabet - Language writing type

  An alphabet is a standardized set of basic written symbols or graphemes that represent the phonemes of certain spoken languages.