Gemini 4 Hydrogen becomes default AI Mode model at $1.50/$7.50 per million tokens

Google has released the everyday tier of its Gemini 4 family. Gemini 4 Hydrogen is optimized for speed in coding, agentic execution and general use, and it is now the default model in AI Mode in Google Search globally and available without a subscription in the free Gemini app. Demis Hassabis promoted it as 'incredible speed and intelligence for everyday coding, agentic work and all-round use'.

Pricing is the main point. At $1.50 per million input tokens and $7.50 per million output, Hydrogen matches midrange competitors on performance while costing roughly a third as much to run, according to Google's materials. It sits below Gemini 4 Carbon, built for deep reasoning, and Gemini 4 Argon, both announced as part of the same rollout and Google's November Search core update. Making Hydrogen the AI Mode default puts a strong agentic model in front of Google's entire search audience, which no competitor can match for distribution.

It also caps a heavy week of Google shipping. Sundar Pichai listed new 8th-gen TPUs on Monday, Gemini 4 Carbon and Gemini Robotics 2.5 on Tuesday, AlphaGenome 2 and the Gemini Decisions API on Wednesday, and Hydrogen on Thursday. The midrange tier is where volume competition is: DeepSeek's V4.1-Flash, xAI's Grok 4.7 at $2/$6 and Meta's new Muse Ember all target the same cheap, fast agent workloads. On HN, Hydrogen's coding focus drew interest even as the NVIDIA news dominated.

Caveats: the 'one-third the cost' claim comes from Google and will need independent benchmarks to confirm, and per-token prices don't always reflect cost per completed task, which is the point of an AWS guide on routing coding agents published this week. Watch independent coding evaluations and whether developers move agent workloads from Anthropic or OpenAI midrange models to Hydrogen.