Google has begun a broad rollout of Gemini 4 Argon, its latest frontier model, to developers and paying subscribers. The model is available through the Gemini API and AI Studio, in the Gemini app and AI Mode in Search for AI Pro and Ultra subscribers, and in Antigravity for Ultra subscribers. Introductory API pricing is $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off; prices double after November 30.
Google says Argon outperforms other frontier models on benchmarks spanning coding, finance, legal, cybersecurity, and science, and cites internal results including a 35% increase in agentic code submissions, a 45% drop in YouTube comment server load, and a 2.85x inference speedup for Gemma models. The Verge reports Google's claim that Argon beats GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5, while Engadget cites Artificial Analysis finding Argon matches GPT-6 Astra's Intelligence Index at 60% of the cost per task. The two outlets thus present different pictures of how Argon compares to rivals.
The rollout is limited to paying tiers. Ars Technica reports that free users and lower-tier subscribers won't get full access, and Antigravity access requires the Ultra plan. The Verge says it is unclear when or if Argon will reach Plus or free users. Google says it is keeping Gemini 3.1 Pro available for those who need a capable model below Argon.
Robotics is one area where Google says Argon shows a 'measurable step increase' in embodied reasoning. A DeepMind robotics specification shows Argon driving a robot arm through a closed-loop task: reorienting a mug upright and placing a napkin inside it, using wrist-camera images, coordinate grids, and motion primitives such as grasp, place, and move_until_contact. That suggests the model's spatial reasoning improvements extend beyond code to physical action.