Google launches Gemini 4 Hydrogen with access for free users
The AMW Read
A widely distributed, lower-priced Gemini model updates Google's foundation-model offering and could affect model selection across the segment.
Google launches Gemini 4 Hydrogen with access for free users
Google announced Gemini 4 Hydrogen on November 12 and began a phased global rollout. The model targets everyday coding and agent tasks, with speed and efficiency as priorities. It accepts text, image, audio, and video inputs and generates text. Developers can access it through Google AI Studio and all Gemini API tiers. In the Gemini app, free users can access it through automatic model selection.
Hydrogen broadens Google's Gemini 4 distribution after the more limited initial availability of Argon and Carbon. Google also makes Hydrogen available in Search and enterprise products. Its API costs $1.50 per million input tokens and $7.50 per million output tokens, regardless of context length. The model supports a one-million-token context window. Google says it completes tasks about 2.5 times faster than Gemini 3.8 Flash, although that claim does not establish results for every workload.
For builders, the practical decision is whether Hydrogen's price, speed, and quality fit their tasks. Google offers four reasoning levels, from low to xhigh, with high as the default. In Google's comparison, Hydrogen led seven of ten benchmark categories, while Anthropic's Claude Sonnet 5.5 led three, including terminal coding and computer use. Google's model card also reports a 5.57-percentage-point increase in excessive refusals versus Gemini 3.8 Flash, especially for prompts containing images.
