正文 · AI 翻译
IT Home reported on November 13 that Google has officially launched the Gemini 4 Hydrogen model, designed for fast execution and efficient operation in everyday coding and agent workflows, bringing the capabilities of the Gemini 4 series to users at every tier.
Input: $1.50 per million tokens (IT Home note: approximately 10.0 yuan at the current exchange rate).
Output: $7.50 per million tokens (approximately 50.1 yuan at the current exchange rate).
This pricing applies across the full 1 million-token context window, with no tiered pricing. On the Artificial Analysis Intelligence Index, Hydrogen scores 51 at the default high thinking level, above Gemini 4 Argon's 50, and reaches 54 at the highest xhigh level. It also often completes tasks faster than models in the same class.
Google also introduced Orchestrator, an intelligent scheduling feature built into the Google Antigravity platform that can automatically switch between models such as Carbon and Hydrogen within the same session. In the Terminal-Bench 4.0 test, this scheduling approach achieved a task success rate close to Carbon's while using 24% less quota on average and completing tasks 19% faster.
Compared with the previous-generation Gemini 3.8 Flash, Hydrogen completes tasks 2.5 times faster at its default high thinking level. Its Terminal-Bench 4.1 score rose from 18.2% to 57%, and its hallucination rate fell by more than a factor of 5.
The model offers four thinking-depth settings—low, medium, high (the default), and xhigh—allowing users to adjust reasoning depth, latency, and cost according to the task. It supports up to 1 million tokens of native input and 1 million tokens of output, and can process multimodal content including large PDFs, spreadsheets, charts, audio, and video.
On safety, Hydrogen reduces agents' actions beyond their authorized scope by up to 81%, inadequate-service issues by 79%, and inaccurate behavior and shortcut-taking in tool-use and coding scenarios by a factor of 12. It scores 96.7% on an internal agent honesty evaluation, achieves a 99.9% chain-of-thought retention rate, and improves resilience against indirect prompt injection on the Gray Swan IPI benchmark by more than 2.3 times.
Starting on launch day, Hydrogen is being rolled out globally: developers can access it through the paid tiers of the Gemini API, AI Studio, and Android Studio; enterprises can deploy it through Gemini for Business, Google Antigravity, and Gemini in Sheets; and general users can select the model from a drop-down menu in the Gemini app. Free users can also receive suitable Hydrogen capabilities through automatic model selection, while Google Search is also incorporating the upgraded model experience.