Overview
- Google has released Gemini 4 Hydrogen and begun rolling it out across developer, enterprise, and consumer channels with access via the Gemini API, AI Studio, Android Studio paid tier, Gemini for Business, Google Sheets, the Gemini app, and Google Search.
- Hydrogen natively supports up to 1,000,000 input and 1,000,000 output tokens and can process large multimodal files such as PDFs, tables, charts, audio, and video while offering four selectable reasoning depths (low, medium, high, xhigh).
- Google published per‑million‑token pricing that covers the full 1M‑token context window, charging USD 1.50 per million input tokens and USD 7.50 per million output tokens with no segmented tiering.
- The Antigravity platform’s new Orchestrator scheduler automatically routes work between models (for example Carbon and Hydrogen) and, in Google’s Terminal‑Bench tests, cut average quota use by 24% and raised completion speed by 19% versus single‑model runs.
- Google supplied comparative safety and performance results showing large gains over Gemini 3.8 Flash — including a reported 2.5× speed boost at default settings, a Terminal‑Bench score jump to 57% from 18.2%, and more than a fivefold drop in hallucinations — but these figures are company‑reported and await independent verification while likely changing cost and design choices for long‑context and multimodal workflows.