By Hamza Faisal

Google has launched Gemini 4 Hydrogen, a fast, low-cost addition to its Gemini 4 family designed for everyday coding and agentic workflows. Announced on November 12 by Google DeepMind, Hydrogen is the workhorse of the lineup: the model that handles high-volume jobs such as writing code, calling tools, reading documents and running agents, while its bigger siblings, Carbon and Argon, take on the hardest reasoning tasks.

The pricing is the headline. Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, flat across its entire one-million-token context window, with no surcharge for long inputs. Google says it takes over from Gemini 3.8 Flash as the default model for high-volume work, and at its default high thinking level it finishes tasks 2.5 times faster than Flash.

Built for speed, priced for volume

Hydrogen ships with four thinking levels — low, medium, high and a new xhigh tier — so developers can dial the reasoning depth up or down for each request, paying and waiting only for what a task needs. At its default high setting, Google reports it scores 51 on the Artificial Analysis Intelligence Index, ahead of the previous flagship-tier Argon at 50, and reaches 54 at the xhigh setting. It also ranks fifth on Agent Arena and leads its class on several specialist benchmarks, including 58.8 percent on HLE-Diamond and top scores on tax and legal research tests. All of these figures are Google's own, and none had been independently replicated at launch.

The model pairs a one-million-token unified context window with support for function calling, code execution, file search and Google Search grounding, and it can turn dense PDFs, spreadsheets, charts, audio and recorded video into cited reports. Google's own model card notes the limits too: a March 2026 knowledge cutoff, occasional hallucinations and slowdowns, and higher token use at the deeper thinking levels.

One model to plan, one to execute

The bigger idea is multi-model teamwork. Alongside Hydrogen, Google is introducing Orchestrator in Antigravity, its agent coding platform, which routes work between models automatically. In parallel-agent workflows, Carbon can act as the coordinator while Hydrogen powers worker and verifier agents in isolated workspaces, building and testing changes at the same time. Google reports Orchestrator reached near-Carbon task success on Terminal-Bench 4.0 while using 24 percent less quota and finishing 19 percent faster on average.

Hydrogen is also now the default agent in Gemini Managed Agents inside Antigravity, and Google has shipped a gemini-api-dev skill to help coding agents use the API. The pitch is straightforward: let the expensive model think, and let Hydrogen do.

Rolling out everywhere today

Hydrogen is rolling out globally from November 12 across all tiers of Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Google Search, with paid Google AI Pro and Ultra subscribers getting it in Antigravity too. Free users of the Gemini app will get Hydrogen through automatic model selection rather than picking it directly.

The launch lands in a week when every major lab is splitting its lineup into planners and workers: Anthropic's Claude Fable 5.2 topped the same independent index days earlier, and Meta shipped a ten-cent model for repetitive agent tasks. Google's answer is to make the workhorse cheap enough to run all day.