What happened
Google’s fastest Gemini 4 model is now free, and it is 2.5 times faster than the model it replaces. Gemini 4 Hydrogen became generally available on 12 November 2026 as the fast, everyday member of the Gemini 4 family, next to the heavier Carbon and Argon models.
Hydrogen is the recommended upgrade from Gemini 3.8 Flash for everyday coding and agent work. Google says it finishes tasks 2.5 times faster than 3.8 Flash, scores 51 on the Artificial Analysis Intelligence Index at its default setting, one point above Argon, and 54 at its new xhigh thinking level. It keeps a 1 million-token context window, with four thinking levels: low, medium, high as the default, and xhigh.
It is free in Google Search for everyone, free with rate limits in Google AI Studio and the Gemini API free tier, and free in the Gemini app, with one catch worth knowing. In the Gemini app, free users get Hydrogen through automatic model selection, when Google decides the prompt fits it; picking it directly from the model menu needs a paid plan. On the paid API it costs $1.50 per million input tokens and $7.50 per million output tokens, flat across the full 1 million-token window, with no dated price increase announced. Carbon and Argon, by contrast, cost $2/$10 and rise to $4/$20 from 1 December.
Hydrogen also landed the same day in Cursor, where it scored 54.3% on CursorBench, Cursor’s own coding benchmark, up from 39.6% for 3.8 Flash. Google says that in Antigravity’s Orchestrator, Hydrogen reaches near-Carbon task success while using 24% less quota and finishing 19% faster.
Image: Google (official Gemini 4 Hydrogen announcement)
Why it matters for developers
The fast model is where your agent loops live, and the fast model just got much better for free. Every retry, every small fix, every “try again with this error” runs on the cheap tier, so a 2.5× speed jump and a benchmark jump from 39.6% to 54.3% on CursorBench is not a rounding error. It is the difference between a fast model you tolerate and a fast model you actually trust with real edits.
The price matters too. At $1.50/$7.50 per million tokens, Hydrogen is the cheapest Gemini 4 model, and it is the only one of the three with a free tier. If you have been paying Carbon prices for jobs Hydrogen can do, you are now overpaying for those jobs.
Nothing breaks today. Gemini 3.8 Flash is still listed as stable, and nothing you built on it stops working. Hydrogen is an upgrade you can switch to, not a deadline you have to meet.
What changes for Claude Code / Cursor / Codex users
- Cursor: the biggest change. Hydrogen is in Cursor’s model list now, at 54.3% on CursorBench. Make it your default fast model and keep Carbon for the hard jobs.
- Claude Code: nothing changes directly. Claude Code runs Claude, not Gemini. But if you use Gemini alongside Claude for cheap bulk work, Hydrogen is now the cheap bulk model to reach for.
- Codex: nothing changes directly. The lesson travels, though: the “fast but dumb” tier is catching up, so check whether your own fast tier still earns its place.
What to watch next
- 1 Dec 2026: Carbon and Argon prices double to $4/$20 per million tokens. Hydrogen’s flat $1.50/$7.50 suddenly looks even better.
- 1 Jan 2027: Gemini 3.8 Flash’s rate doubles to exactly Hydrogen’s price. After that date there is no price reason left to stay on Flash.
- Your own agent loops: try Hydrogen as the build model with a heavier model planning, and measure the result before you commit your whole workflow to it.