The short version
Gemini 4 Hydrogen is Google’s new low-cost execution model: cheaper than Carbon, faster than the Flash model it replaces, and available on every Gemini API tier. Google pitches it for writing code, running tests and acting as the worker inside multi-model agent setups. Teams doing AI agent development should test it as the default model for routine steps and keep frontier models for the hard ones.
What did Google release on November 12?
Gemini 4 Hydrogen is the fast, inexpensive member of the Gemini 4 family. Google describes Carbon and Argon as its models for deep, long-horizon reasoning, and Hydrogen as the one built for rapid execution across agent workflows. TechCrunch notes the logic behind the launch: the most powerful model is not always needed, and a faster, cheaper, reasonably smart one often does the job.
Hydrogen is rolling out globally on these surfaces:
- Developers: every tier of Google AI Studio and the Gemini API, plus Android Studio.
- Enterprises: Gemini for business.
- Antigravity: Google’s agentic development platform, for paid AI Pro and Ultra subscribers only.
- Consumers: the Gemini app and Google Search, where The Verge reports it is the new default model in AI Mode.
How fast is Gemini 4 Hydrogen, and how good?
Google’s headline claim is speed: tasks finish 2.5 times faster than on Gemini 3.8 Flash at the default “high” thinking level. The quality claims are just as large, so it helps to see who is making each one.
| Claim | What the sources report |
|---|---|
| Coding in a terminal | Google: 57% on Terminal-Bench 4.1, versus 18.2% for Gemini 3.8 Flash |
| General intelligence | Google, confirmed by TechCrunch: 51 on the Artificial Analysis Intelligence Index at the default level, one point above Gemini 4 Argon; 54 at xhigh |
| Agent tasks | Google and TechCrunch: No. 5 on Agent Arena for deep research, data analysis and codebase navigation |
| Accuracy | Google: more than 5× fewer fabricated claims than 3.8 Flash |
| Partner testing | Google: 2× to 2.8× faster runs and 12% to 23% fewer tokens than 3.8 Flash on enterprise workflows |
| Peers | The Verge, citing Google: matches or beats some models in its class, such as Claude Sonnet 5.5 and GPT-6.1 Terra |
Google also says Hydrogen edits code surgically, without rewriting unrelated files, and reads large repositories by batching file reads and searches in parallel.
Its security figures are encouraging: 2.3 times more resilient to indirect prompt injection on the Gray Swan benchmark, and up to half as likely to bypass permission boundaries on blocked tasks. Those results come from vendor testing.
How much does Gemini 4 Hydrogen cost?
Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, with no surcharge for long prompts anywhere in its 1 million-token window. It can also return up to 1 million output tokens.
For comparison, Google lists Gemini 4 Carbon at $2 / $10 until November 30 and $4 / $20 after that. At the standard rates Hydrogen costs well under half as much as Carbon per token.
As always, per-token price is not per-task cost. Hydrogen’s four thinking levels (low, medium, high and xhigh) change how many tokens a task burns, so the setting you choose matters as much as the rate card.
What is Orchestrator in Antigravity?
Orchestrator is a new Antigravity feature that switches between Gemini models inside one session instead of using a single model from start to finish. Google says that on Terminal-Bench 4.0 it reached near-Carbon success rates while using 24% less quota and finishing 19% faster on average.
For parallel work, Google describes a split of roles. Carbon acts as the coordinator. Hydrogen powers the worker and verifier agents that build and test changes in isolated workspaces.
What it means for US & EU software teams
Hydrogen makes the planner-and-worker pattern cheaper to run. Most agent steps are routine: read a file, apply a small patch, run the tests, check the result. Running those steps on a frontier model at frontier rates is hard to justify, and a fast model that scores 57% on a terminal coding benchmark changes the math for teams building coding assistants, internal dev tools or support agents.
The risk is the same one that comes with every cheap default: quality drifts on the edge cases you did not test. Treat Google’s numbers as a reason to run your own evaluation, not as its result. Keep a fixed set of real tasks, record success rate and tokens per completed task, and compare Hydrogen with your current default at more than one thinking level.
For EU teams, the usual checks apply before any customer or regulated data goes to a new model. Confirm where Gemini API and Gemini for business process and store data, and which terms cover your tier.
What should teams do this week?
- Pick the routine steps. List the agent steps that rarely fail today, such as small edits, test runs and lookups, and trial Hydrogen there first.
- Test two thinking levels. Run the same tasks at high and medium, and compare success rate, latency and tokens.
- Escalate, don’t replace. Route tasks that fail on Hydrogen to Carbon or your current frontier model automatically.
- Keep the guardrails. Sandboxes, least-privilege credentials and human review before merge stay in place, whatever the injection scores say.
Frequently asked questions
What is Gemini 4 Hydrogen?
Gemini 4 Hydrogen is a fast, low-cost model in Google’s Gemini 4 family, released on November 12, 2026. Google built it for everyday coding and agent workflows, such as writing code, running tests and verifying changes, while Gemini 4 Carbon and Argon handle deep, long-horizon reasoning.
How much does the Gemini 4 Hydrogen API cost?
Google prices Hydrogen at $1.50 per million input tokens and $7.50 per million output tokens, flat across its full 1 million-token context window. Gemini 4 Carbon costs $2 and $10 until November 30, 2026, then $4 and $20.
Where can developers use Gemini 4 Hydrogen?
It is available on all tiers of Google AI Studio and the Gemini API, in Android Studio and in Gemini for business. In Google Antigravity it is limited to paid Google AI Pro and Ultra subscribers. It also powers the Gemini app and AI Mode in Google Search.
Is Gemini 4 Hydrogen better than Gemini 4 Argon?
On one broad benchmark, yes. Google says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default thinking level, one point above Argon, and TechCrunch reported the same comparison. Google still positions Argon and Carbon for the hardest reasoning tasks, so test both on your own workload.
Sources
Google — Introducing Gemini 4 Hydrogen (primary source, November 12, 2026)
Google for Developers — Gemini 4 Hydrogen: everyday coding and agentic workflows (November 12, 2026)
TechCrunch — Google’s latest Search upgrade promises better answers to your most complicated questions (November 12, 2026)
The Verge — Google’s new Gemini model is better at understanding complicated questions (November 12, 2026)