Google announced Gemini 4 Hydrogen on Nov. 12, positioning it as a fast model for everyday coding and agentic workflows. The company says Hydrogen is rolling out in Google AI Studio and the Gemini API, with additional availability in Android Studio’s paid tier, Google Antigravity for paid AI Pro and Ultra subscribers, Gemini for business, the Gemini app and Google Search. This is a named, public release, replacing the earlier rumor status that surrounded “Hydrogen.”
What will developers pay?
Google’s AI Studio pricing page lists Hydrogen’s standard Gemini API rates as $1.50 per million input tokens and $7.50 per million output tokens on the paid tier. The listed free tier has no token charge, subject to Google’s model and usage limits. Output pricing includes thinking tokens. Google also says every processed token counts, including internal tool instructions and function declarations when grounding is used, so real requests may cost more than a prompt-length estimate suggests.
The same pricing page lists batch and flex options at lower token rates and a priority tier at higher rates. It shows a unified context window of up to one million tokens and long-decode generation of up to one million output tokens. Those figures describe API limits, not a guarantee that each request will be accepted at maximum size or that output will arrive synchronously; the model card and API documentation should be checked for the exact request path.
Where can you access it?
Google’s release says developers can call Hydrogen through the Gemini API and try it in AI Studio. It also says Android Studio access is on a paid tier. Antigravity access requires a paid Google AI Pro or Ultra subscription. Consumer access in the Gemini app and Search is a separate product rollout; it does not mean every feature or quota is identical to the API. Developers should select the intended surface and check its billing, rate limits, and data-use terms separately.
Google’s announcement describes Hydrogen as multimodal and says it can accept text, images, audio and video. It reports a one-million-token context and up to one million output tokens through long-decode generation. The public model card provides additional input and output details, while the live pricing documentation is the place to confirm current rates and tier availability before release.
How strong is it compared with Argon?
Google says Hydrogen scored 51 on the Artificial Analysis Intelligence Index at its default high thinking level, compared with 50 for Gemini 4 Argon, and reached 54 at xhigh. The announcement also describes selected improvements in speed, coding and tool use. These are Google’s reported figures and comparisons; they are not a guarantee that Hydrogen will outperform Argon on a particular app or workload. VIBPS has not run an independent benchmark.
The model choice depends on the task. Google positions Carbon and Argon for deeper, long-horizon reasoning, while Hydrogen is aimed at faster execution and everyday agent work. Its post also describes Antigravity’s Orchestrator switching between Hydrogen and frontier models, and reports lower quota use and completion time for that product-specific setup. Teams should treat those savings as Google’s reported evaluation and compare cost, latency and success rate on their own tasks.
For developers, the useful update is that a previously unannounced model now has a published model ID, live API pricing and broad product availability. Before switching production traffic, test the model with representative prompts, measure thinking-token costs, verify output limits and set a fallback for workflows that depend on a particular reasoning depth. VIBPS did not independently test the model.
How Google’s Gemini Decisions API handles structured classification
VIBPS is the social network for software, where makers can publish applications built with new model APIs and reach people who want to follow their work.