Google retires its Flash line in favour of Gemini 4 Hydrogen, billed as faster and less prone to invention
The lighter model becomes the default in Google Search and works alongside heavier models in the Antigravity coding environment.
Google on Thursday released Gemini 4 Hydrogen, the successor to its faster and cheaper Flash series. It immediately replaces Gemini 3.8 Flash across the company's products, Ars Technica writes. According to Google, Hydrogen runs two and a half times faster than its predecessor even at a high thinking level and makes up facts more than five times less often. On the Artificial Analysis Intelligence Index it scores 51 in its default mode and 54 in its new top setting, against 50 for the heavier Gemini 4 Argon, 9to5Google reports. The model can take up to one million tokens of input.
In Google Search, Hydrogen becomes the default model for AI Mode, with more visual answers and more links to sources. In the Gemini app, paying AI Pro and Ultra subscribers get the model straight away. Reports differ on free users: Ars Technica writes that they keep an older model in the app for now, while 9to5Google says they get Hydrogen whenever automatic model selection deems it suitable. Developers pay 1.50 dollars per million input tokens and 7.50 dollars per million output tokens through the API, the same as for 3.8 Flash. The heavier Carbon and Argon models will, however, become more expensive next month, according to Ars Technica.
Google is also adding Orchestrator to its Antigravity coding environment. A planning model such as Carbon or Argon first explores the code and draws up a plan, after which Hydrogen edits files and runs tests, InfoWorld reports. If things repeatedly go wrong, the planning model takes over again. Analysts quoted by InfoWorld see it as a way to cut token costs, but warn that companies must be able to trace which model made which code change.