Google introduced Google Gemini 4 Hydrogen today, a new model built for speed and efficiency in everyday coding and agentic workflows, and made it available across Google AI Studio, the Gemini API, Gemini for business, the Gemini app, Google Search, and paid Google AI Pro and Ultra subscribers in Google Antigravity.
The launch gives developers and business users a model that Google says is meant to run quickly and cheaply without giving up capability. Hydrogen is priced at $1.50 per million input tokens and $7.50 per million output tokens across a 1 million-token context window, and Google said it can serve as a standalone model for everyday work or as the execution layer alongside Gemini 4 Carbon and Gemini 4 Argon in multi-model setups. SVP said, “Hydrogen brings cutting-edge intelligence to everyone.”
That is the reason the release lands now: Google is not just adding another name to the lineup, but putting a new option in front of users on day one. For teams building code tools, agents, and search-driven workflows, the appeal is straightforward. Hydrogen is meant to handle the fast, repetitive parts of the job, while the heavier reasoning models stay focused on longer, more complex tasks. Google also said the model can be used across a 1 million-token context window, which gives it room for larger projects and longer sessions.
The numbers make the pitch sharper, but they also create a contradiction worth noting. Google describes Hydrogen as the high-efficiency engine for rapid execution, yet it also said the model surpassed Gemini 4 Argon on the Artificial Analysis Intelligence Index at its default high thinking level, with a score of 51 versus 50. At its highest xhigh thinking setting, Hydrogen reached 54. In other words, the model is being sold as the fast and lower-cost choice, but Google is also leaning on a benchmark result that puts it ahead of one of its frontier models, at least in that test.
Google pushed that logic further inside Google Antigravity, where it introduced Orchestrator to switch between models like Gemini 4 Carbon and Gemini 4 Hydrogen in the same session. On Terminal-Bench 4.0, Orchestrator reached near-Carbon task success rates while using 24% less quota on average and finishing 19% faster. Google also said Hydrogen finished tasks 2.5 times faster than Gemini 3.8 Flash at its high default setting, more than tripled Terminal-Bench 4.1 scores to 57% from 18.2%, cut fabricated claims by more than 5x, and adapted cleanly when a tool call failed or a search came back empty.
Those details point to what Google wants Hydrogen to become: the model that handles speed-sensitive work without forcing users to move between products or pay for heavier reasoning when they do not need it. The open question is how far Google will extend that rollout beyond the products and tiers named in the announcement, and whether developers will treat Hydrogen as the default execution engine or keep it as a specialist tool for fast-turn work.
