Key Notes
- Google released Gemini 4 Hydrogen for everyday coding, research and document work, with support for text, images, audio and video.
- Google reports a Terminal-Bench 4.1 score of 57%, compared with 18.2% for Gemini 3.8 Flash, while claiming faster task completion.
- API pricing is $1.50 per million input tokens and $7.50 per million output tokens, and Antigravity can pair Hydrogen with larger models through Orchestrator.
Google has released Gemini 4 Hydrogen, a model designed to handle everyday coding, research and document work quickly, while taking on execution tasks assigned by the company’s more powerful AI systems.
The company announced the model on November 12 and began rolling it out across its developer tools, business products, Gemini app and Search. API pricing starts at $1.50 per million input tokens and $7.50 per million output tokens.
The launch broadens Google’s Gemini 4 lineup beyond frontier reasoning. Carbon and Argon can handle difficult planning decisions; Hydrogen is meant to keep routine work moving. That division becomes more useful as AI products run longer tasks involving code changes, tests, searches and repeated checks.
A faster successor to Flash for daily work
Google says Hydrogen completes tasks 2.5 times faster than Gemini 3.8 Flash at its default high thinking setting. On Terminal-Bench 4.1, it reports a score of 57%, compared with 18.2% for Flash.
Those are Google’s reported results on particular evaluations. They point to a substantial improvement, but do not establish that every coding task will finish 2.5 times faster or that every generated change will be correct.
Independent evaluator Artificial Analysis scores Hydrogen at 51 on its Intelligence Index at the high setting. Its assessment places the model above the average of comparable offerings for intelligence and task speed, while noting relatively heavy output-token use. That matters because a quick model can still accumulate costs if it produces long responses or repeats work.
Carbon plans, Hydrogen executes
The accompanying Orchestrator feature in Google Antigravity switches between models as a task develops. Carbon can inspect a repository and work out a plan, then hand implementation and testing to Hydrogen. If the work runs into difficulty, control can return to Carbon to reconsider the approach.
The distinction is more flexible than selecting one model at the beginning and leaving it in charge throughout. A routine dependency update may turn into a difficult debugging problem; an architectural decision may be followed by dozens of straightforward edits.
Google’s Antigravity team says Orchestrator achieved task success close to Carbon alone while using 24% less quota and finishing 19% faster in its testing. These are company-reported averages, rather than a fixed discount on every project.
Orchestrator with Carbon or Argon is rolling out to Google AI Ultra subscribers, while Pro subscribers get the Argon option. Hydrogen itself is available in Antigravity for paid Pro and Ultra users. Developers can also assign models to specific roles through the Antigravity SDK.
The release follows the Carbon expansion that AIstify covered earlier this week, alongside the preview of parallel agent teams. Hydrogen adds a faster execution option to that broader strategy.
Documents, images and video fit into the same workflow
Google’s model card specifies support for text, images, audio and video, within a unified context window of up to one million tokens. Hydrogen is adapted from Carbon and is intended for repository debugging, multimodal knowledge work, web development and other agent tasks.
Four thinking levels, low, medium, high and xhigh, let developers vary the reasoning effort. High is the default. More reasoning can improve an answer, but may also increase waiting time and token consumption.
The model card also lists familiar limitations: fabricated answers remain possible, and requests can occasionally slow down or time out. Its stated knowledge cutoff is March 2026, so up-to-date research still depends on access to current sources.
Broad availability, with different access rules
Hydrogen is reaching AI Studio, the Gemini API, Gemini for business and consumer products. In the Gemini app, free users receive it through automatic model selection when Google considers it appropriate for the request. That is different from unrestricted manual access to every model and thinking level.
For developers, the practical question is how much work can be completed at an acceptable cost. Hydrogen’s launch gives them another choice: use it independently for everyday tasks, or let a larger model delegate the steps that need speed more than extended deliberation.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
