Google launched Gemini 4 Hydrogen on November 12, bringing the model to all Search users globally while access in other products varies by subscription and account. Free Gemini app users receive Hydrogen through automatic model selection when Google considers it appropriate for a prompt; Antigravity requires Google AI Pro or Ultra, according to Google's launch announcement.

Ars Technica reporter Ryan Whitwam reported that Hydrogen replaces Gemini 3.8 Flash across Google's products, with access beginning immediately. iPhone in Canada reporter Austin Blake also reported the global Search rollout. The launch spans everyday assistance, coding and agent workflows.

Gemini 4 Hydrogen Access Varies by Product

Google is also bringing Hydrogen to AI Studio, the Gemini API and Gemini for business. Gemini in Sheets starts with Pro and Ultra consumer subscriptions and eligible business subscriptions. Access to Hydrogen in Search therefore does not establish access to the same model in a spreadsheet or coding workspace.

In Search, Google says Hydrogen improves complex-question handling and produces more visual answers. Its upgraded query fan-out conducts simultaneous searches across subtopics to surface more relevant websites. Language improvements extend across supported AI Mode languages, the company says in its Search announcement.

Account authorization is a separate consideration: Google's proposed administrator-managed Search and AI access for K-12 accounts included a selected-school pilot planned for later November, as reported in our coverage of proposed school Search and AI controls.

Coding Speed Comes With Company Benchmarks

Google positions Hydrogen for everyday coding and agent workflows, reserving Carbon and Argon for longer reasoning tasks. It claims Hydrogen completes tasks 2.5 times faster than Gemini 3.8 Flash and reduces fabricated claims more than fivefold. The claimed reduction in fabricated claims comes from Google's analysis, Ars Technica reported.

Google also introduced Orchestrator, which switches models within an Antigravity session. For developers weighing coding-agent workspaces, that makes model selection another workflow consideration: different tasks within one session can use different models.

Google Cloud recommends the latest Google Gen AI SDK for Hydrogen integrations. Hydrogen supports up to 65,536 output tokens in a foreground segment; longer generation uses continuation or background decoding. Google Cloud also classifies it as a short-term generally available release with a minimum 45-day retirement notice policy. That lifecycle matters when planning maintenance for an integration.

Pricing and Reliability Shape the Trade-Off

Google lists standard Hydrogen pricing at $1.50 per million input tokens and $7.50 per million output tokens. API usage therefore has a separate cost calculation from the subscription requirements governing access in Antigravity and Sheets.

Hydrogen is adapted from Gemini 4 Carbon and has a shared input-output context window, according to Google DeepMind's November model card. Hydrogen can still hallucinate, experience timeouts and use more tokens at higher thinking levels, Google DeepMind says. Its stated knowledge cutoff is March 2026.

Hydrogen has a unified 1,048,576-token budget covering input, thinking, function calls and final output, according to Google Cloud's technical guide. It offers LOW, MEDIUM, HIGH and XHIGH thinking settings, defaults to HIGH and does not support explicit context caching.

For coding and agent workflows, those characteristics put speed, token consumption and reliability in the same calculation. Google claims faster task completion, while Google DeepMind acknowledges that Hydrogen can still hallucinate and experience timeouts.

Photo: Gemini 4 Flown Silver Fliteline Medallion by Commissioned by NASA astronaut office, unknown manufacturer (attributed to "Fliteline") (Public domain), via Wikimedia Commons.