Gemini API: Gemini 4 Hydrogen Reaches General Availability with a New xhigh Thinking Level
Google released Gemini 4 Hydrogen (gemini-4-hydrogen) to general availability in the Gemini API on November 12, 2026. The model targets fast, everyday coding and agentic workflows. It is available on the free tier and costs $1.50 per 1M input tokens and $7.50 per 1M output tokens on the paid tier. It adds a fourth xhigh thinking level, and it is now the default model for the Antigravity agent in Gemini Managed Agents.
Key Takeaways
- Gemini 4 Hydrogen reached general availability on November 12, 2026, and it is available on the free tier as well as the paid tier.
- Paid pricing is $1.50 per 1M input tokens and $7.50 per 1M output tokens, flat across the full 1M context window, with cached input at $0.075 per 1M tokens.
- A new
xhighthinking level joinslow,mediumand the defaulthigh, giving deeper reasoning for hard multi-step problems at the cost of latency and tokens. - Google says Hydrogen cuts fabricated claims by more than 5x versus Gemini 3.8 Flash by flagging missing context instead of guessing, and it batches read and search tool calls in parallel.
- Hydrogen is now the default model for the Antigravity agent in Gemini Managed Agents.
- Migrating from Gemini 3.8 Flash has breaking changes: no explicit caching, sampling parameters removed,
minimalthinking rejected, and a unified 1,048,576-token window with a 65,536-token synchronous output ceiling.
Gemini 4 Hydrogen Is Now Generally Available
Google released Gemini 4 Hydrogen (gemini-4-hydrogen) to general availability in the Gemini API on November 12, 2026. Google positions the model as bringing cutting-edge intelligence to everyone, with speed and efficiency for everyday coding and agentic workflows. It is available on both the free and paid tiers in Google AI Studio and the Gemini API.
Hydrogen joins Gemini 4 Carbon and Gemini 4 Argon in the Gemini 4 family. Like those models, it has a unified 1M-token context window and long decode of up to 1M tokens.
Pricing and Availability
On the paid tier, Gemini 4 Hydrogen costs $1.50 per 1M input tokens and $7.50 per 1M output tokens, with output pricing including thinking tokens. Pricing is flat across the full 1M context window. Cached input tokens cost $0.075 per 1M tokens through implicit caching, a 95% discount on cache hits. Implicit caching runs by default on prompts of at least 4,096 tokens.
The model is also free of charge on the free tier, subject to rate limits. Google counts and bills every token the model processes, including thinking tokens and internal system instructions and function declarations used with grounding.
A Fourth Thinking Level: xhigh
Hydrogen supports four thinking levels: low, medium, high (the default) and the new xhigh.
lowuses the fewest tokens and has the lowest latency. It suits cost- and latency-sensitive workloads such as fast chat and high-throughput processing.mediumuses fewer tokens and responds faster. It suits interactive coding, live UI prototyping and latency-sensitive tool calls.highbalances intelligence and speed, and Google recommends it as the default for most tasks.xhighuses the most tokens for the deepest reasoning. Google advises reserving it for hard, multi-step coding and agentic problems wherehighfalls short. It raises latency and token use and can cause overthinking on simpler prompts.
Hydrogen defaults to high rather than medium and rejects the minimal level with an error.
What Google Says Is New in the Model
Google highlights several areas of improvement:
- Fast agentic execution. Hydrogen writes cleaner code and cuts fabricated claims by more than 5x compared with Gemini 3.8 Flash by flagging missing context instead of guessing. It also batches read and search tool calls in parallel and adapts better when tool calls fail.
- Multi-model orchestration. The model works standalone, or pairs with Carbon and Argon to power parallel worker and verifier agents at lower latency without driving up token cost.
- Multimodal reasoning. The model reads tables, charts, dense PDFs, images, audio and video with higher accuracy. It supports agentic video understanding to inspect specific frames and visual details.
- Unified context and long decode. Input, thinking and output tokens share one 1,048,576-token window. Synchronous outputs run up to 65,536 tokens, and longer outputs use background execution or client-side continuation.
Default Model for the Antigravity Agent
The Antigravity agent in Gemini Managed Agents now uses Gemini 4 Hydrogen by default, because of its improved performance and reasoning.
Migrating From Gemini 3.8 Flash
Google calls Hydrogen the recommended upgrade from Gemini 3.8 Flash. It uses the same Gemini 4 API design as Carbon and Argon, and the migration checklist covers these changes:
- Change the model string from
gemini-3.8-flashtogemini-4-hydrogen. - Remove hardcoded low
max_output_tokenscaps or raise them to 65,536, since input, thinking and output now share one window. - Remove
cached_contentfields from requests. Hydrogen does not support explicit caching (cachedContentsandcaches.create), only implicit caching. - Remove
temperature,top_pandtop_k. The server ignores custom values if they are sent. - Declare tools natively with function calling, or constrain JSON with structured outputs.
- Preserve conversation state and thought signatures across turns.
Teams already running Carbon or Argon need no request schema changes to add Hydrogen. Hydrogen supports up to 65,536 tokens per non-streaming hop, compared with 32,768 tokens per non-streaming REST hop on Carbon and Argon. Planning turns can be routed to Carbon and execution turns to Hydrogen within one multi-turn session.