Google on November 12, 2026 introduced Gemini 4 Hydrogen, a generally available model that the company says is built for speed and for everyday coding and agentic work, rather than the long-horizon reasoning of Gemini 4 Carbon and Gemini 4 Argon. The announcement was written by Koray Kavukcuoglu, senior vice president at Google DeepMind and chief AI architect at Google, and appeared on Google's blog and the Google DeepMind blog. The Gemini API release notes list the model as gemini-4-hydrogen, now generally available, including on the free tier, with paid-tier pricing of $1.50 per million input tokens and $7.50 per million output tokens.
Google says Hydrogen is priced flat across a one-million-token context window and can emit up to one million output tokens. It adds a fourth thinking level, called xhigh, alongside low, medium and the default high setting. In the API notes, Hydrogen becomes the default model for the Antigravity agent in Gemini Managed Agents. Kavukcuoglu writes that the model is rolling out across Google AI Studio, the Gemini API, Gemini for business, the Gemini app, Google Search, and Google Antigravity for paid Google AI Pro and Ultra subscribers. In the Gemini app, free users receive Hydrogen through automatic model selection when Google judges it the right model for the prompt. A companion Google post says Search is being upgraded with Hydrogen for users globally.
The company positions Hydrogen as both a standalone everyday model and a partner to Carbon and Argon in multi-model workloads. With the release, Google is introducing Orchestrator in Google Antigravity, which it says can switch between models such as Carbon and Hydrogen inside the same session. Google reports that on Terminal-Bench 4.0, Orchestrator reached near-Carbon task success rates while using 24 percent less quota on average and finishing 19 percent faster. Those figures come from Google and have not been independently audited here.
Google also compares Hydrogen with Gemini 3.8 Flash, the earlier fast model that many developers would be leaving. It says Hydrogen finishes tasks 2.5 times faster at its default high setting, more than triples Terminal-Bench 4.1 scores (57 percent versus 18.2 percent), and cuts fabricated claims by more than five times. Other scores in the launch post, all reported by Google, include a class lead on HLE-Diamond at 58.8 percent, a first-place 52.3 percent on the Vals Tax Agent Bench, a tie for first at 55.8 percent on the Vals Legal Research Bench, 53.4 percent on the Multimodal Knowledge Agent benchmark, and 71.6 percent on Agents' Last Exam v1.1. Google also says Hydrogen ranks fifth on Agent Arena, the live benchmark from arena.ai. Readers should treat these as vendor-reported results.
On the consumer side, Google says Hydrogen follows instructions more closely in the Gemini app and helps Search understand messy or multi-part queries, with more visual responses and a wider set of links through an upgraded query fan-out method. The separate Search post presents this as a major upgrade, but it belongs to the same November 12 model launch rather than a distinct product announcement.
Google describes safety work aimed at a faster model used in everyday agentic tasks, including claimed reductions in overreach and under-service, stronger resistance to indirect prompt injection on the Gray Swan IPI benchmark, and more precise refusals that try to separate biosecurity risks from legitimate medical questions. Those evaluations are Google's own. The launch post points readers to a Gemini 4 Hydrogen model card for more detail. What is firmly established is the public availability and the official price: a new Gemini 4 model is generally available, named and priced by Google, and is being wired into Search, the Gemini app, the API and Antigravity.