Gemini 4 Hydrogen
Overview
Gemini 4 Hydrogen is a Google DeepMind model in the Gemini 4 family, released on November 12, 2026. Adapted from Gemini 4 Carbon, it focuses on efficient everyday coding, agent execution, and multimodal knowledge work, with adjustable reasoning and a unified 1M-token context window.
Key Features
- Iterative Coding: Supports repository exploration, issue resolution, debugging, and targeted code changes.
- Agent Execution: Handles multi-step tool workflows and can serve as a worker or verifier alongside other models.
- Multimodal Analysis: Combines text, documents, images, charts, audio, and recorded video for research and synthesis.
- Adjustable Reasoning: Offers four thinking levels to balance reasoning depth, response time, and token use.
- Extended Generation: Long decode supports lengthy reports, code, and reasoning within the remaining context budget.
Best Use Cases
- Software Maintenance: Investigating bugs, answering codebase questions, implementing fixes, and checking changes with tests.
- Document-Based Research: Synthesizing reports, tables, charts, and recorded material into summaries or structured findings.
- Web Prototyping: Building interactive applications and refining interfaces through repeated implementation and review.
- Worker and Verifier Agents: Executing defined subtasks, gathering evidence, and checking results in coordinated agent workflows.
Capabilities and Limitations
| Capability | Description |
|---|---|
| Reasoning | Supports low, medium, high (default), and xhigh thinking levels. |
| Multimodal | Accepts text, images, audio, video, and PDFs; produces text. |
| Document Formats | Google's Cloud API also accepts Word, PowerPoint, Excel, and OpenDocument files. |
| Context Window | 1,048,576 tokens shared across input, thinking, function calls, and final output. |
| Generation Limit | 65,536 combined thinking, function-call, and final-output tokens per default synchronous segment; long decode can use the remaining unified context budget through continuation or background processing. |
| Tool Use | Supports function calling, code execution, URL context, grounding, and computer use through compatible integrations. |
| Structured Outputs | Supports schema-constrained responses for extraction and structured deliverables. |
| Video Understanding | Agentic video processing is supported but requires enabling; static processing is the default. |
| Prompt Caching | Google's API applies implicit caching to shared prompt prefixes of at least 4,096 tokens; explicit caching is unsupported. |
| Knowledge Cutoff | March 2026; knowledge in some domains may be less current. |
Known Limitations
- Shared Generation Budget: Thinking and function calls consume output tokens. Higher thinking levels can increase token use and latency, leaving less room for the final answer under a fixed limit.
- Output Length: Exact word-count or token-length targets are unreliable.
- Generation Controls: Custom
temperature,top_p, andtop_kvalues are ignored. Theminimalthinking level and explicit thinking budgets are unsupported. - Reliability: Hallucinations, occasional slow responses, and timeouts remain possible; factual conclusions and generated code require verification.
Standard Pricing
The token prices below are shown in USD per 1 million tokens; web search is billed in USD per use.
| Model | Input (USD / 1M Tokens) | Cache Write (USD / 1M Tokens) | Cache Read (USD / 1M Tokens) | Output (USD / 1M Tokens) | Web Search (USD / use) |
|---|---|---|---|---|---|
| Gemini 4 Hydrogen | $1.50 | $1.50 | $0.075 | $7.50 | $0.014 |
Credits settlement: B.AI converts charges at 1 USD = 1,000,000 Credits and deducts Credits from the account balance.
Web search costs 14,000 Credits per use, equivalent to $0.014 per use.
For this model on B.AI, cache writes are priced at 1x the input rate. This is a B.AI billing rule, not a separate cache-write price listed by Google. The model supports implicit caching only; the cache-write price does not imply support for explicit cache creation. Google's implicit caching requires prompts of at least 4,096 tokens; meeting that threshold does not guarantee a cache hit on every request.
Model prices in the documentation are based on the model providers' official list prices and serve as B.AI standard reference prices. B.AI may offer discounts or limited-time promotions on this basis; offer details are shown in Promotions and Pricing Updates and the website's discount groups. Applicable prices, offer terms, and final charges are subject to the platform display and actual billing records.