Google has introduced Gemini 4 Hydrogen, a new AI model designed to make everyday coding, software development, and agentic workflows faster and more efficient. Positioned as the more execution-focused member of the Gemini 4 family, Hydrogen aims to bring advanced AI capabilities to a wider audience without requiring users to rely on Google’s more computationally demanding models for routine tasks.

Announced on November 12, Gemini 4 Hydrogen joins the company’s existing Gemini 4 Carbon and Gemini 4 Argon models. While Carbon and Argon are designed for complex reasoning and longer-running tasks, Hydrogen focuses on rapid execution, making it particularly useful for writing code, debugging projects, running tests, and handling multi-step tasks through AI agents.

Google says Hydrogen can also outperform Argon on the Artificial Analysis Intelligence Index at its default High thinking level, scoring 51 compared with Argon’s 50. The model reaches a score of 54 when running at its highest XHigh thinking setting.

Built for faster coding and agentic execution

One of Gemini 4 Hydrogen’s biggest priorities is improving the speed and reliability of everyday software development.

The model can work through existing code repositories, identify problems, debug software, and make targeted changes without unnecessarily rewriting unrelated files. It can also answer questions about large codebases and handle iterative development tasks that require repeated testing and refinement.

Google says Hydrogen is particularly effective at exploring large repositories because it can batch file-reading and search operations in parallel instead of inspecting files one at a time.

Compared with Gemini 3.8 Flash, the company reports that Hydrogen completes tasks 2.5 times faster at its default High thinking setting. It also achieves a Terminal-Bench 4.1 score of 57%, compared with 18.2% for its predecessor.

Google further claims that the model reduces fabricated claims by more than five times and handles unsuccessful tool calls and empty search results more effectively.

Beyond traditional coding, Hydrogen is designed for broader agentic workflows, including research, data analysis, and navigating complex projects. It ranks fifth on Agent Arena, according to Google’s announcement, while delivering strong results across several specialized benchmarks.

Google wants Hydrogen working alongside its bigger models

Rather than positioning Hydrogen as a replacement for its more powerful models, Google sees it as an efficient execution partner.

The company is introducing Orchestrator in Google Antigravity, a feature that dynamically switches between Gemini models during the same session. This allows the system to use Carbon for complex planning and architectural decisions before handing execution-heavy tasks to Hydrogen.

For example, Carbon can coordinate a development project while Hydrogen-powered worker and verifier agents build, test, and review changes simultaneously in isolated workspaces.

Google says this approach achieved near-Carbon task success rates on Terminal-Bench 4.0 while consuming 24% less quota and completing tasks 19% faster on average.

This multi-model approach could be particularly useful for developers working on larger projects. Instead of spending expensive reasoning tokens on every routine coding operation, teams can reserve their more capable models for decisions that genuinely require deeper reasoning.

Four thinking levels offer more control

Gemini 4 Hydrogen includes four thinking levels: Low, Medium, High, and XHigh.

High is the default setting, while XHigh provides additional reasoning capacity for tasks that need more careful analysis. Users can adjust the level depending on whether they prioritize response speed, reasoning quality, or computational cost.

The model also supports multimodal workloads. It can process text, images, audio, and video, making it useful for working with different types of information rather than limiting interactions to conventional text prompts.

Google highlights its ability to analyze lengthy PDFs, spreadsheets, charts, audio recordings, and recorded videos, producing structured reports and other deliverables.

Hydrogen supports a unified context window of up to 1 million tokens. According to its model card, synchronous responses support up to 64,000 output tokens by default, while background long decoding can extend output to 1 million tokens.

More than just a coding model

Although software development is a major focus, Google is also bringing Hydrogen to its consumer products.

In the Gemini app, the model is designed to follow instructions more closely, use the appropriate tools, and complete everyday tasks with fewer follow-up prompts. This includes writing, research, document preparation, product comparisons, and summarizing information from multiple sources.

It can also connect information from uploaded files and Google services such as Gmail and Google Drive, alongside supported connected apps including Airtable and Dropbox.

Google Search is receiving an upgrade powered by Hydrogen as well. The company says the model improves how Search understands complicated queries, including questions containing multiple requirements, typos, or loosely structured thoughts.

Search responses are also getting more visual presentations and a wider range of relevant links through an improved query fan-out technique.

For developers and creators, Hydrogen can also generate interactive web experiences, prototypes, and applications with fewer prompts. Google demonstrated this capability with several projects, including an interactive 3D globe powered by real-time lightning strike data, a WebGL-based visualizer, and an e-commerce storefront featuring interactive 3D elements.

The model can also work with Google Sheets and Slides, including generating animated spreadsheet visuals and creating presentations with automated layout checks.

Pricing, availability, and supported platforms

Google says Gemini 4 Hydrogen is rolling out globally across its developer, enterprise, and consumer products.

Developers can access the model through the Gemini API and Google AI Studio, including all AI Studio tiers. It is also available through Android Studio.

For enterprise customers, Hydrogen is accessible through Gemini for business, Google Antigravity, and Gemini in Google Sheets for eligible subscriptions.

Consumers can access it through the Gemini app and Google Search. Google AI Pro and Ultra subscribers can also use it in Google Antigravity, while Gemini in Google Sheets is available to eligible Pro and Ultra subscribers.

Free Gemini app users will receive access through automatic model selection when Hydrogen is considered appropriate for their prompts.

For API users, Google has set pricing at $1.50 per million input tokens and $7.50 per million output tokens. The pricing remains the same across the model’s entire 1-million-token context window.

Google’s faster model comes with important limitations

Google’s Gemini 4 Hydrogen model card also outlines some limitations worth keeping in mind. Like other generative AI systems, the model can produce inaccurate information, and occasional slowdowns or timeouts may occur.

The company also acknowledges that higher thinking levels can consume additional tokens as the model works to improve its responses. Its stated knowledge cutoff is March 2026, although information freshness can vary between subject areas.

Google says it has strengthened safeguards against indirect prompt injection attacks, inappropriate tool usage, and attempts to bypass permission boundaries. However, these protections do not eliminate the need for users to review AI-generated results and supervise actions involving external tools.

With Gemini 4 Hydrogen, Google is taking a practical approach to its latest AI model family. Not every task requires the deepest possible reasoning, particularly when developers are repeatedly writing code, testing changes, or working through everyday requests.

By combining faster execution, configurable thinking levels, multimodal capabilities, and collaboration with its more advanced models, Hydrogen gives Google another way to bring Gemini 4 capabilities into everyday products and professional workflows.