Google on November 12 introduced Gemini 4 Hydrogen, a new tier of its Gemini 4 family positioned not as the smartest model but as the cheapest fast one — the "execution engine" for everyday coding and agentic work.

Where Google's frontier models Carbon and Argon are pitched at long-horizon reasoning, Hydrogen is built to write code, run tests, verify changes and call tools quickly and repeatedly. Google says it surpasses Argon on the Artificial Analysis Intelligence Index at its default "high" thinking level (51 versus 50, rising to 54 at a new "xhigh" setting) — a claim that puts a cheap model at or above the company's own flagship on a widely cited aggregate.

Pricing is the headline. Hydrogen is priced flat across its entire 1-million-token context window at $1.50 per million input tokens and $7.50 per million output tokens, with no long-context surcharge.

Google is also shipping Orchestrator in its Antigravity developer environment, which switches dynamically between models — for example Carbon for architecture and Hydrogen for execution — inside a single session. On Terminal-Bench 4.0, Google says Orchestrator reached near-Carbon task success while consuming 24% less quota and finishing 19% faster. Hydrogen also triples the Terminal-Bench 4.1 scores of the model it replaces, Gemini 3.8 Flash (57% versus 18.2%), finishes tasks 2.5 times faster and cuts fabricated claims by more than five-fold, per Google's figures. The company reports a 96.7% "honesty rate" on an internal agent-honesty evaluation and says the model is more than 2.3 times more resilient to indirect prompt injection on the Gray Swan benchmark.

Availability is broad from day one: the Gemini API and AI Studio, the paid tier of Android Studio, Gemini for business, Google Sheets for eligible subscriptions, the Gemini app, Google Search for all users, and Google's Antigravity IDE for paid AI Pro and Ultra subscribers. Search gets upgraded intent handling, more visual answers and a wider query fan-out; the app follows multi-turn instructions more closely.

Caveats: the benchmark figures are Google's own or from vendor-friendly leaderboards, and the multi-model efficiency numbers come from partner testing during an early-access period. The concrete, checkable claims are the pricing and the context-window terms. Developers migrating from Gemini 3.8 Flash should consult the model card and migration guide for API differences.