Google Antigravity has launched Gemini 4 Hydrogen alongside Orchestrator, a system that pairs Hydrogen with Gemini 4 Carbon in a dynamic, bidirectional workflow. Instead of relying on static model routers, Orchestrator delegates tasks like planning and refactoring to Carbon while routing execution steps like edits and testing to Hydrogen. On Terminal-Bench 4.1 testing, this coordinated setup matches near-Carbon success levels while finishing 19% faster and utilizing 24% less quota. Gemini 4 Hydrogen is also available as a standalone coding model featuring four reasoning depth settings ranging from low to xhigh.
Key points
Developers can use the Antigravity SDK to manually build custom pipelines dividing tasks…: Developers can use the Antigravity SDK to manually build custom pipelines dividing tasks between Hydrogen and Carbon.
Paid Google AI Pro and Ultra subscribers receive access to Gemini 4 Hydrogen starting tod…: Paid Google AI Pro and Ultra subscribers receive access to Gemini 4 Hydrogen starting today.
Orchestrator is initially restricted to Google AI Ultra users, with broader tier availabi…: Orchestrator is initially restricted to Google AI Ultra users, with broader tier availability and enterprise access planned.
Sources checked by our analysis pipeline: 1
Explore our next generation products
Built for developers and teams in the agent-first era
Everything you need to stay up-to-date and get help
Gemini 4 Hydrogen and Orchestrator
Today, we’re rolling out Gemini 4 Hydrogen inside Google Antigravity alongside a new model dropdown option called Orchestrator, which pairs Hydrogen with Gemini 4 Carbon in an optimized balance of quality, efficiency, and latency.
Put Gemini 4 Carbon and Gemini 4 Hydrogen to work with Orchestrator
Earlier this week, we introduced Gemini 4 Carbon alongside /swarm
for tackling massive, multi-worktree engineering tasks with parallel agent teams. In just a couple of days, tens of thousands of swarms have been launched on Antigravity, pushing frontier intelligence on the hardest problems.
For everyday feature work and debugging inside a single session, running a frontier reasoning model with multiple parallel agents on every step of every routine task slows down your loop and burns frontier quota. A common industry approach is to use a static “model router” that guesses task difficulty from your first prompt and routes the entire task to the cheapest model that…
Small businesses in Singapore are deploying artificial intelligence to overcome budget constraints, reduce production timelines, and tackle complex initiatives previously reserved for larger corporations. Recognition from national industry bodies showcases how practical machine learning applications are enabling lean teams to secure international partnerships and streamline routine operations. These technological adoptions allow enterprises across various sectors to scale their capacity while curbing substantial a…
Emerging corporate models show that integrating artificial intelligence requires rigorous human oversight to catch errors and maintain high professional standards. As small businesses adopt automation, organizations are implementing internal mentorship programs and secondary verification teams rather than relying solely on automated outputs. These operational adjustments ensure companies benefit from machine processing power while mitigating critical risks in regulatory compliance and customer operations.
Consumer advocacy organizations are demanding an immediate emergency freeze on next-generation computing systems, arguing the technology poses uncontrollable risks to public welfare. Filed in San Francisco Superior Court, the civil complaint seeks to mandate rigorous safety standards before tech leaders can train or deploy further models. This legal push aims to force industry giants to prove human control while compelling government intervention on emerging digital threats.