Google released Gemini 4 Hydrogen on November 12, targeting coding, debugging, and agent execution. API pricing is $1.50 per million input tokens and $7.50 per million output tokens, with a million-token context window.
Hydrogen offers four reasoning levels: low, medium, high, and xhigh. High is the default. Google positions the model for rapid execution alongside Carbon and Argon, which handle longer reasoning tasks.
Google also introduced Antigravity Orchestrator. The feature switches models within a session rather than assigning the entire job to one model. Google reports 24% lower quota consumption and 19% faster completion on Terminal-Bench 4.0, with task success approaching Carbon. These figures describe Google’s evaluation.
The model accepts text, images, audio, and video. Its model card specifies a unified input-output context window of up to one million tokens. Default synchronous output is limited to 64,000 tokens, with longer output supported through background decoding.
Distribution includes the Gemini API, AI Studio, Gemini app, Search, business products, and Google Antigravity. Access conditions differ across those services.
The model card lists hallucinations, occasional timeouts, and increased token consumption at higher reasoning levels among known limitations. Its knowledge cutoff is March 2026, with less recent coverage in some subjects. Google also reports increased unjustified refusals against Gemini 3.8 Flash in its automated safety evaluation, particularly for multimodal prompts.
Comments (0)