Google releases Gemini 4 Argon AI model for developers and enterprises

Google has made its latest frontier AI model, Gemini 4 Argon, globally available to developers, enterprises, scientists, and consumers. The model is accessible via the API in AI Studio and the Gemini platform, as well as through Google AI Ultra and Pro subscriptions in the Gemini app and AI Mode in Search. According to a blog post by Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect at Google, the model is designed for "deep work in complex, real-world tasks across domains including coding, finance, legal, cybersecurity, and science."

Performance and Pricing

Google states that Gemini 4 Argon ranks #1 on the Vals Index, a benchmark measuring model performance across economic sectors such as finance, coding, legal, and tax work. The model launches with an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%.

In internal engineering tests, Google reports that Argon helped the Chrome team complete an architectural refactor across hundreds of files in weeks, a task that typically takes months. The company also claims the model increased agentic code submissions by more than 35% while reducing code rollbacks by nearly a third. Additionally, Google says Argon achieved a 2.85x end-to-end inference speedup on an A100 for its open-source Gemma models.

Industry and Scientific Applications

Early access partners have reported specific outcomes using the model. Alex Cabrera, Founding Engineer at Axiom Bio, stated, "On our Scientific Agent Benchmark, Gemini 4 was at the frontier across most tasks, including our hardest: ranking the clinical risk between two drugs." Gopal Srinivasan, Global AI and Data Lead at Deloitte, noted that in their evaluations, the model "completed successfully 100% of the time" for long-running tasks.

In the medical field, Brendan Fortuner, SVP Technology at Ambience Healthcare, reported that Gemini 4 "achieved the highest recall and ICD-10 code specificity among the models we evaluated" on complex synthetic longitudinal charts. In legal tech, Caio Perona, CEO of MinutaIA, said the model "actually surpassed our human expert's answer key by surfacing a legally relevant detail we had missed" in one evaluation.

Google also highlighted scientific breakthroughs, including the use of Argon to help discover a broader class of counterexamples to "Grad’s conjecture" in nuclear fusion research, alongside Professor Matt Landreman of the University of Maryland. In enzyme design, Google scientists reported that Argon identified five novel designs for serine esterases in a matter of hours, with subsequent laboratory testing confirming robust catalytic activity.

Robotics and Physical Reasoning

A dedicated post on the Google Developers Blog by Kanishka Rao, Principal Software Engineer, details Argon's capabilities in robotics. The model scores 87% on the Embodied Reasoning (ERQA) benchmark, which tests the ability to solve physical, spatial problems in 3D environments. Google says this score outperforms other frontier models.

The blog post describes several applications, including:

• Robot Control: Argon controlled a Boston Dynamics Spot quadruped to learn how to step on a foot-pedal trash bin by first practicing in a MuJoCo simulation.
• Real2Sim: The model can reconstruct physically calibrated simulation twins from robot camera feeds or human video.
• Self-Improvement: In one test, Argon generated simulation data that boosted a real-world robot's success rate from 56% to 75% and reduced completion time from 35s to 24s.

Why it matters

The release of Gemini 4 Argon represents a significant expansion of Google's AI capabilities into high-stakes professional and scientific domains. By providing specific metrics on coding efficiency, legal analysis, and scientific discovery, Google is positioning the model as a tool for complex, multi-step reasoning rather than just conversational interaction. The integration of advanced physical reasoning capabilities also signals a push toward more autonomous robotics applications, potentially accelerating development in fields requiring precise physical manipulation and simulation.

---

Sources

• blog.google: Gemini 4 Argon is now available for everyone
• developers.googleblog.com: Gemini 4 Argon for Robotics - Google Developers Blog

This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.