Gemini 4 Hydrogen and Orchestrator

Today, we’re rolling out Gemini 4 Hydrogen inner Google Antigravity alongside a new example dropdown choice called Orchestrator, which pairs Hydrogen alongside Gemini 4 Carbon in an optimized balance of quality, efficiency, and latency.

Put Gemini 4 Carbon and Gemini 4 Hydrogen to activity alongside Orchestrator

Earlier this week, we introduced Gemini 4 Carbon alongside /swarm for tackling massive, multi-worktree engineering tasks alongside parallel delegate teams. In fair a brace of days, tens of thousands of swarms have been launched on Antigravity, pushing frontier intellect on the hardest problems.

For mundane characteristic activity and debugging inner a sole session, operating a frontier reasoning example alongside multiple parallel agents on all stage of all regular project slows downward your iteration and burns frontier quota. A average industry method is to use a fixed “model router” that guesses project difficulty from your archetypal immediate and routes the complete project to the cheapest example that is fine adequate to complete the task. Unfortunately, these commencement breaking downward as agents rotate into longer running, requiring distinct levels of complexity at distinct stages, frequently discovered as the project evolves.

Orchestrator takes a dynamic, bidirectional method between Carbon and Gemini 4 Hydrogen inside the development of a sole trajectory. For example, Carbon can commencement in a read-only phase to examine dependencies, determine ambiguities, and build a tangible scheme alongside definitive re-planning triggers. Hydrogen could afterward obtain complete to compose edits, update packages, and run tests. If Hydrogen hits friction, it would hand rear to Carbon to inspect the state, repair the plan, and resume execution. This allows Orchestrator to attain near-Carbon project achievement during being faster and additional quota-efficient. On Terminal-Bench 4.1, Orchestrator reached near-Carbon project achievement rates while, on average, consuming 24% small quota and completing 19% faster.

• Explicit multi-model pipelines alongside the Antigravity SDK: When you desire programmatic authority complete which example handles all role, you can cable Hydrogen and Carbon together immediately alongside the Antigravity SDK. In the demo below, an open-source matter triage scheme built alongside the Antigravity SDK pairs multiple agents together. The Hydrogen agents grip triaging and coding to keep up alongside incoming matter volume, searching the codebase, asking clarifying questions, and accelerated UI repeat from PR comments during the Carbon delegate is assigned to larger refactors, new features, and safety checks.

• Fast, dependable implementation alongside Gemini 4 Hydrogen: Hydrogen additionally plant as a standalone example for mundane coding and agentic work, scoring 51 on the Artificial Analysis Intelligence Index (surpassing Argon’s 50, and reaching 54 at its highest setting) and 57% on Terminal-Bench 4.1 (more than tripiling 3.8 Flash’s 18.2%). When selecting Hydrogen directly, you can dial reasoning degree throughout four thinking levels, low, medium, elevated (default), and a new xhigh tier.

Getting started

Gemini 4 Hydrogen brings cutting-edge intellect for mundane coding and agentic workflows. When you desire frontier preparedness paired alongside speed without managing example switches, Orchestrator puts the two Gemini 4 Carbon and Gemini 4 Hydrogen rearward one prompt. Starting today, Gemini 4 Hydrogen is rolling out throughout paid Google AI Pro and Ultra subscribers. Orchestrator is rolling out for Google AI Ultra subscribers, alongside additional models and plans supported in the future. Enterprises can admission the two in Gemini for business.