Get Gemini 4 Hydrogen Free With a 1M-Token Context Window

Feed it 1,048,576 tokens per request, get 65,536 back, with four thinking levels.

Hydrogen is the fast model in Google's Gemini 4 family, built for code, tests and tool calls; Carbon and Argon handle deeper reasoning. Google calls it the upgrade from Gemini 3.8 Flash.

The catch: free-tier prompts train Google's products, the per-model rate limits are unpublished, and Batch and Search grounding stay paid-only.

How to claim: 1) Open Google AI Studio → 2) create a project and an API key → 3) call gemini-4-hydrogen through the Interactions or GenerateContent API.

What changed

Released Nov 12, 2026, gemini-4-hydrogen runs on both the free and paid tiers of the Gemini API and AI Studio, through the Interactions and GenerateContent APIs. Free-tier input, output and implicit caching are listed as "free of charge"; Batch and Search grounding are not.

The model card gives a unified 1,048,576-token window, 65,536 tokens of synchronous output by default and up to 1M with long decode. Paid pricing is flat across that window: $1.50 per 1M input tokens and $7.50 per 1M output, half on Batch.

Google says it runs 2.5x faster than Gemini 3.8 Flash at default high thinking and lifts Terminal-Bench 4.1 from 18.2% to 57%. Free Gemini app users reach it through automatic model selection.