THE ESSENTIALS IN 30 SECONDS

  • Gemini 4 Hydrogen launched on November 12, 2026 for developers, businesses and consumers at once, with no closed early-access phase.
  • API pricing is $1.50 per million input tokens and $7.50 per million output tokens, flat across a 1-million-token context window.
  • Google says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default setting, one point above Gemini 4 Argon; the benchmark figures are Google's own.

Google has added a third model to its Gemini 4 line. Gemini 4 Hydrogen, announced on November 12, 2026 in a post signed by Koray Kavukcuoglu, is pitched as a fast, lower-cost option for routine coding and agent tasks, while Gemini 4 Carbon and Gemini 4 Argon remain the company's frontier models for long, demanding reasoning.

Where it is available and what it costs

Unlike Argon and Carbon, which SiliconANGLE notes were first offered to vetted cybersecurity defenders in Google's Fairwind Program, Hydrogen went straight to general use. Google says it is rolling out worldwide in the Gemini API and Google AI Studio across all tiers, in Android Studio, in Gemini Enterprise and in Google Antigravity for AI Pro and Ultra subscribers. Consumers can pick it in the Gemini app, where free users may also be routed to it automatically, and it is being used in Google Search.

Developers pay $1.50 per million input tokens and $7.50 per million output tokens, a flat rate over the full 1-million-token context window. For comparison, SiliconANGLE reports that Argon costs $2 and $10 under introductory pricing that ends on November 30, rising to $4 and $20. Hydrogen also introduces a fourth reasoning level, called xhigh, above the default high setting.

Google's performance claims

According to Google, Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default setting, against 50 for Argon, and 54 at xhigh. The company says it finishes tasks 2.5 times faster than Gemini 3.8 Flash and scores 57% on Terminal-Bench 4.1, versus 18.2% for that earlier model. Google also lists strong results on tax, legal research and multimodal agent benchmarks, plus internal safety tests that it says show fewer fabricated claims and fewer actions beyond what users asked.

In Antigravity, a new Orchestrator feature switches between Carbon and Hydrogen within one session, for instance letting Carbon plan while Hydrogen runs worker agents. Google says this came close to Carbon's success rate on Terminal-Bench 4.0 while using about 24% less quota.

What to keep in mind

All of these figures come from Google or partners it selected; SiliconANGLE's coverage relays them without independent testing. The lead over Argon on the Artificial Analysis index is a single point. Google's post gives no regional exclusions beyond describing a global rollout, and some features, such as Gemini in Google Sheets, start with paid subscribers.