Gemini 4 Hydrogen is live — and it’s Google’s new speed-and-value play in the AI model race. Launched globally today, it replaces Gemini 3.8 Flash across Google Search, the Gemini app, Google AI Studio, the Gemini API, Gemini for business, and paid tiers of Google AI Pro and Ultra in Google Antigravity.
No new hardware required. No waitlist. Just faster responses, tiered reasoning, and a price tag that undercuts rivals on cost-per-intelligence, at least according to Google’s own benchmarks and third-party index data cited by Wccftech. It runs 2.5× faster than Gemini 3.8 Flash on everyday tasks, coding, and agentic workflows, Sundar Pichai confirmed.
That speed comes with four distinct “thinking levels”, low, medium, high, and xhigh, letting users choose how deeply the model reasons before generating output. You pay only for the level you use. At the xhigh level, Gemini 4 Hydrogen scores 54 on the Artificial Analysis Intelligence Index, one point shy of Claude Sonnet 5.5 (max) and GPT-6.1 Sol (max), and four points ahead of Gemini 4 Argon (high).
But its real edge is economics: $1.50 per million input tokens and $7.50 per million output tokens. At xhigh, Google says it costs roughly 32% as much as Claude Sonnet 5.5 (max) for near-equivalent intelligence, while costing about 2.5× more than GPT-6.1 Sol (max) on the same task. Output speed hits ~200 tokens per second.
And per Vals AI’s latest Top 6 ranking, which now includes only Anthropic and Google models, Hydrogen is both the cheapest and fastest. OpenAI’s GPT-6 Astra dropped to eighth. This isn’t a limited preview or a regional rollout. It’s live everywhere Google AI services operate, including India and the US, and available to all existing subscribers and developers without upgrade prompts or waiting periods.
What’s next? The model is already active in Search and the Gemini app. Developers can start tuning it via AI Studio and the API today. Business customers get it through Gemini for business.
And Google Antigravity, the company’s enterprise AI infrastructure layer, now serves Hydrogen as its default flash-class model for latency-sensitive workloads. No launch event. No keynote slide. Just a quiet, global switch, and a clear signal: speed and cost efficiency are now table stakes, not differentiators.
CDFiled by The Computing Desk
Laptops, desktops, processors, graphics cards and the operating systems that run on them.
More from this desk →
Be the first to comment
Join the argument. No password, just your email or a passkey.