The Google presented, on Wednesday (30), the Gemini 4 Argon, new artificial intelligence (AI) model that now leads the Gemini 4 line. According to the company, it is their most advanced model so far, developed to handle complex and long-duration tasks in areas such as programming, cybersecurity, and professional activities.
The launch occurs after months of waiting for the successor of the previous generation of Google's advanced models. Argon initially reaches a selected group of cybersecurity partners, while the company conducts security evaluations before potential wider availability. Google has not yet informed when the model will be publicly available.
According to the company, the Gemini 4 Argon set a new record in real software development tasks and achieved first place in cybersecurity evaluations. The model also presented the best results in another test focused on professional tasks in financial, legal, and other corporate areas.
Model is larger and targets professional tasks
- According to Google, Argon is larger than the previous generation's Pro models and was developed to perform tasks that require multiple steps and greater reasoning capacity;
- The company claims that the model shows comparable performance to frontier systems, such as the Astra from OpenAI, and the Opus from Anthropic, in programming and cybersecurity evaluations;
- Beyond these areas, Google points to applications in finance, law, science, and mathematics;
- The company also stated that the model has been used internally by quantum computing researchers.
The new model is not only being prepared for external users. According to Google, Argon is already used internally to optimize memory usage in its data centers.
The company claims that the application allowed to free up hundreds of terabytes of memory without the need to purchase additional hardware.
According to Google, Argon is larger than the previous Pro models and was developed to handle tasks that require multiple steps and greater reasoning capacity – Image: Mijansk786/Shutterstock
Read more:
- Google challenges EU court orders to share data with European search and AI rivals
- Gemini now answers the phone for you — and you can take over at any time
- Gemini gets a face: new Google AI speaks, reacts, and synchronizes lips in real time
Before making the Gemini 4 Argon widely available, Google decided to subject it to security testing and evaluation.
Initially, the model will be offered to selected cybersecurity partners. The company also participates in a voluntary program by the United States government for early access to AI models before the public launch.Google announced that it is conducting a pre-launch security assessment with the U.S. government. The company also aims to strengthen protections against various forms of misuse, including known attacks such as prompt injection, before expanding access to the model.
Security concerns have gained prominence precisely at a time of heightened attention to the risks of AI systems capable of performing tasks more autonomously.Gemini 4 arrives after months of waiting.
The launch of Argon occurs after a prolonged period without a new flagship model from Google.The company had been working on a version called Gemini 3.5 Pro, but ended up putting the project on hold before its launch. Reuters reported that the development took longer than expected and that the model stopped being competitive compared to rival systems.
With the Gemini 4 Argon, Google returns to focusing its strategy on a high-performance model capable of competing directly with the most advanced systems of other AI companies.
The arrival of the new model also occurs a day after Google's CEO, Sundar Pichai, signed a voluntary AI safety agreement with U.S. President Donald Trump during a meeting with executives from major technology companies at the White House.
Rodrigo Mozelli is a journalist graduated from the Universidade Metodista de São Paulo (UMESP) and is currently a writer at Olhar Digital.
