Google has launched Gemini 4 Hydrogen, a new artificial intelligence model built specifically for fast, low-cost coding and agentic work rather than heavyweight reasoning.
The company said developers can begin building with the model immediately through Google AI Studio, with wider access across its developer and business products. The model was released on November 12, 2026, alongside a separate announcement from Google DeepMind.
What the new model does
Google positions Hydrogen as the workhorse tier beneath its more powerful Gemini 4 Carbon and Argon models, which the company reserves for deep, long horizon reasoning.
In its own announcement, Google said it is “introducing Gemini 4 Hydrogen, bringing cutting-edge intelligence to everyone, with speed and efficiency for everyday coding and agentic workflows.”
The model is designed to handle rapid, iterative work. According to Google, Hydrogen handles repository coding and debugging, interactive web development, and multimodal workflows, running across four thinking levels (low, medium, high, and a new xhigh tier) that scale intelligence, latency, and token use , standalone or alongside frontier models like Gemini 4 Carbon and Gemini 4 Argon.
Google claims Hydrogen finishes tasks 2.5 times faster at its default setting and cuts fabricated claims by more than fivefold compared with the model it replaces.
Pricing and performance claims
Gemini 4 Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens across its entire context window, which Google says is 25% cheaper than the current promotional pricing of Carbon and Argon , though the model costs more than the current promotional pricing of the Gemini 3.8 Flash model it replaces.
Independent coverage of Google’s benchmark data found that Hydrogen’s own benchmark claims are notable, but they come from Google: on Terminal-Bench 4.1, it scored 57 percent, versus 18.2 percent for Gemini 3.8 Flash.
Reporting on the rollout also noted that Google says Hydrogen completes tasks 2.5 times as fast as Gemini 3.8 Flash at Hydrogen’s default reasoning setting , more than tripling its Terminal-Bench 4.1 score over the prior model.
Google’s pricing structure keeps costs flat regardless of how much of the context window is used. Pricing is flat across the full window, and output pricing includes thinking tokens.
The same analysis noted a built-in cost break for developers reusing prompts, since implicit caching cuts input to $0.075 per 1M for prompts of 4,096 tokens or more, a rate that notably matches what Gemini 3.8 Flash’s standard pricing will fall to once its current promotional pricing ends on January 1, 2027.
Where developers can use it
Access to Hydrogen spans Google’s full developer stack. The company said in its announcement that Gemini 4 is being brought to everyone, for all tiers of Google AI Studio, the Gemini API, Gemini for business, the Gemini app, Google Search, and paid Google AI Pro and Ultra subscribers in Google Antigravity.
Enterprises get a similar path, with Google noting the model can be deployed in Gemini for business via the API or model drop down, Google Antigravity, and in Gemini in Google Sheets for eligible business subscriptions.
Inside Google’s Antigravity coding environment, Hydrogen is paired with a new feature called Orchestrator that automatically routes tasks between models depending on difficulty.
One independent summary of that integration explained that on Terminal-Bench 4.0, Orchestrator reached near-Carbon task success rates while consuming 24% less quota and completing tasks 19% faster on average , with Carbon exploring dependencies and building a plan before Hydrogen writes the changes, updates packages and runs the tests.
Why the launch matters
The release reflects a broader shift among AI labs toward offering cheaper, faster models for high-volume, everyday tasks rather than only chasing frontier benchmark scores.
That approach mirrors a trend already reshaping other corners of the coding tool market, where usage-based billing for AI assistants has been climbing as more programmers lean on automated agents for routine work, a dynamic GCN has previously reported on.
It also comes as developer tools increasingly hand AI systems more autonomy to execute code changes directly, a trend GCN has covered in the context of Copilot’s terminal permissions.
Google has not disclosed usage figures for Hydrogen since launch, and the model remains closed source, with access tied to Google’s own API and subscription tiers.
