Gemini 4 Hydrogen Pushes AI Coding Into a Faster Gear

Google has launched Gemini 4 Hydrogen, a new model built to handle coding, research, complex questions, and everyday business work with more speed and capability. It is now rolling out across Google Search, the Gemini app, Google AI Studio, the Gemini API, Google Antigravity, and business products, putting one model across both consumer and developer workflows.

Hydrogen targets the jobs that often slow AI systems down: writing code, fixing errors, running tests, working through documents, and completing multi-step tasks. Google says it can also give more helpful answers to “questions with multiple parts, hard-to-understand typos, or highly specific requests” by breaking each question into parts and addressing them in its response.

A Gemini Model Built for Code and Complex Work

Gemini 4 Hydrogen joins Gemini 4 Carbon and Gemini 4 Argon in Google’s latest model family. Hydrogen focuses on coding and testing, while Carbon can handle complex reasoning, creating a division of labor that Google is using inside its Google Antigravity platform.

That system includes Orchestrator, a feature that automatically switches between models based on task complexity. On Terminal-Bench 4.0, Google says Orchestrator achieved task success rates close to Carbon while using 24% less quota on average and completing tasks 19% faster. The promise is straightforward: use the heavier model when a problem demands deep reasoning, then move routine coding or testing work to Hydrogen.

Hydrogen also brings a unified context window of up to 1 million tokens, covering input and output together. It can process documents, spreadsheets, charts, audio, and recorded video, then create reports and other structured outputs from that material.

The model accepts text, images, audio, and video, while Google Cloud can ingest Office and OpenDocument files directly. The shared window includes reasoning tokens and tool calls, and supported APIs can continue generating past the 65,536-token limit of a single foreground segment.

Speed Gains Meet Tougher Benchmarks

Google says Hydrogen completes tasks 2.5 times faster than Gemini 3.8 Flash at Hydrogen’s default high reasoning level. On Terminal-Bench 4.1, Hydrogen scored 57% compared with 18.2% for Gemini 3.8 Flash, although Google’s comparison table lists Claude Sonnet 5.5 at 63.9%.

Artificial Analysis testing also places Hydrogen ahead of earlier Gemini models. At the default high thinking level, it scored 51 on the Artificial Analysis Intelligence Index, compared with 50 for Gemini 4 Argon. The score rises to 54 at the highest extra-high thinking setting, while Gemini 3.8 Flash scored 31 at high reasoning in the independent testing cited by Google.

The Intelligence Index v5.0.1 combines nine evaluations across areas that include coding, document analysis, and science. Its cost-per-task figure uses a weighted estimate for that evaluation suite, while API rates can vary with prompt length, provider, or time of day.

Google also says Hydrogen recovers more reliably when a tool call fails or a search returns nothing. In an early-access test, Box’s agent reached 79% accuracy on its Complex Work Eval with Hydrogen, up from 74% with Flash. Box CEO Aaron Levie said Hydrogen took less than half as long and used about 12% fewer tokens on a typical task, with the largest gains appearing in work that required reading tables and doing calculations.

Access, Pricing, and Developer Changes

Gemini 4 Hydrogen is rolling out globally across developer, enterprise, and consumer products. Developers can access it through the Gemini API, Google AI Studio, and the paid tier of Android Studio. Enterprise customers can use it through Gemini for business and eligible Google Sheets subscriptions.

The model is also available in the Gemini app and Google Search. Free Gemini app users receive it through automatic model selection when Google determines it fits their prompts, while Google AI Pro and Ultra subscribers can access it through Google Antigravity and Gemini in Sheets.

Robby Stein, Google Search’s vice president of product, called Hydrogen “now the new default model in AI Mode & being used for your hardest questions in AI Overviews.” Google says Search can conduct “simultaneous queries across all the relevant subtopics,” return a wider range of links, and produce responses with images, comparison tables, graphs, and maps. The model also improves language understanding “across all supported AI Mode languages.”

Standard Gemini API pricing is $1.50 per million uncached input tokens, $0.075 per million cached input tokens, and $7.50 per million output tokens, including reasoning tokens. Those rates stay flat across the 1-million-token context window, and Google lists no introductory promotion for Hydrogen’s standard tier.

• Batch, Flex, and Contributor cost half the standard rates.
• Batch processes groups of requests in the background with a target turnaround of 24 hours.
• Flex handles individual requests at lower priority, with a target response time of 1–15 minutes and limited availability when demand is high.
• The invite-only Contributor tier offers standard-tier speed and reliability while allowing Google to use participating request data to improve products and train models.

Developers moving an existing application will need to adjust more than the model name. Hydrogen defaults to high reasoning instead of Flash’s medium setting and adds an “xhigh” option, while it rejects the old thinking_budget parameter and the MINIMAL setting. On Google Cloud, automatic caching starts at 4,096 tokens, down from 6,144 on Flash, and explicit caching is unavailable at launch.

Hydrogen’s rollout connects faster coding, larger multimodal inputs, model selection, and agent workflows in one package. With Google Search already using it for difficult questions and developers gaining access through the API, Google is positioning Gemini 4 Hydrogen as the workhorse model for the next wave of AI-powered software and business tasks.

Stay connected via Google News

Follow us for the latest AI News