
What happened
On November 12, Google released Gemini 4 Hydrogen, a speed-tuned model built on Gemini 4 Carbon, and made it available to all users in the Gemini app, Google Search, the API and Antigravity.
Why it matters
The model is weighing becoming the default for free-tier Gemini app users when Google's auto-selection deems it a fit, and paying Google AI subscribers can pick it explicitly.
WHO IT HITSDevelopers building on the Gemini API and Google AI Studio get a low-cost model for coding, debugging and multi-step research tasks, while Google Workspace and Gemini for business administrators can offer it to enterprise users.
Summaries like this, every morning.
Gemini 4 Hydrogen is the third entry in the Gemini 4 series, following Argon in September and Carbon in October. Google positions Carbon as its highest-performance model, and Hydrogen is tuned from it for speed and efficiency rather than deep reasoning, which the company leaves to Carbon and Argon. That division of labor extends to Google's developer tool Antigravity, where a new Orchestrator feature automatically switches between a top model and Hydrogen within a single session.
The model's intended roles run from focused coding and debugging to multi-step research and knowledge work such as turning PDFs, spreadsheets, audio and video into sourced reports. In a demo, Hydrogen generated a Google Slides deck from a YouTube lecture video and a design template, then checked and fixed overlapping text on its own. Google also says the model makes drafting emails, building study guides and making product comparison tables in the Gemini app more convenient.
On safety, Google's model card compares Hydrogen with the previous fast model, Gemini 3.8 Flash: safety on inputs including images improved, but refusals of legitimate requests also rose, especially for inputs containing images. Google also cautions that response delays and timeouts occasionally occur, and that higher thinking levels can consume many tokens.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
US energy companies including Duke Energy are testing Anthropic's Mythos AI model to identify software vulnera…

Charter's AI Pulse survey of 5,002 full-time U.S

The note defines embedding as a conversion mechanism that maps text and images into numerical vectors while pr…

LVMH Group's Media Division and OpenAI signed a strategic agreement to deploy OpenAI tools and ChatGPT Enterpr…

BOOKERS now supports MCP, so users can connect their account to Claude or ChatGPT (web version, confirmed work…

IMPACTLAKE began offering a ChatGPT plugin for impactlake(TM) and added Model Context Protocol (MCP) support…
