Google Releases Gemini 4 Hydrogen With 1M Context and Antigravity Orchestrator
Google DeepMind launched Gemini 4 Hydrogen at $1.50 per million input tokens alongside Orchestrator routing in Google Antigravity, bringing Gemini 4 intelligence to Search and developers.

By Jin, Founder & Startups
, updated 3 min read
Google on November 12 introduced Gemini 4 Hydrogen, expanding its fourth-generation model family to everyday developer workflows, Google Search, the Gemini app, and enterprise subscribers across Google AI Studio and Google Antigravity.
Positioned as a high-speed execution sibling alongside frontier reasoning models Gemini 4 Carbon and Gemini 4 Argon, Hydrogen pairs a flat 1-million-token context window with substantially lower token costs than flagship tiers. Google DeepMind priced the API at $1.50 per million input tokens and $7.50 per million output tokens, avoiding tiered price increases across long contexts.
"Hydrogen brings cutting-edge intelligence to everyone, with speed and efficiency for everyday coding and agentic workflows," said Koray Kavukcuoglu, senior vice president at Google DeepMind and Google's chief AI architect, in the official announcement.
Benchmark Gains Over Argon and Gemini 3.8 Flash
At its default high thinking setting, Gemini 4 Hydrogen scored 51 on the Artificial Analysis Intelligence Index, stepping ahead of Gemini 4 Argon at 50 and climbing to 54 when switched to its maximum xhigh reasoning effort. Compared with its predecessor Gemini 3.8 Flash, Google reported a 2.5x speedup at the default setting alongside more than triple the score on Terminal-Bench 4.1 (57% versus 18.2%) and over a fivefold drop in fabricated claims.
Across domain-specific evaluations released with the model:
Agent Arena: Ranked #5 overall on Arena.ai's live multi-step agent benchmark.
HLE-Diamond: Reached 58.8% on deep domain reasoning problems.
Enterprise agent tests: Ranked #1 overall on Vals Tax Agent Bench (52.3%) and tied #1 on Vals Legal Research Bench (55.8%).
Antigravity Orchestrator and Gemini Decisions API
Alongside the model weights, Google rolled out Orchestrator inside Google Antigravity for AI Pro and Ultra plans. Instead of locking a coding session to a single model, Orchestrator dynamically switches between Gemini 4 Carbon for architectural planning and Gemini 4 Hydrogen for high-frequency edits, test execution, and verification.
On Terminal-Bench 4.0, Google reported that multi-model Orchestrator runs matched near-Carbon completion rates while consuming 24% less quota and finishing 19% faster. Google also opened developer access to the Gemini Decisions API, designed for low-latency calibrated multimodal routing.
Sources
- 1Google Keyword: Introducing Gemini 4 Hydrogenblog.google
- 2Google Developers: Build With Gemini 4 Hydrogenblog.google