Google launches Gemini 4 Hydrogen, scores 57% on coding benchmark, triples Flash

Google has unveiled Gemini 4 Hydrogen, adding another model to its frontier lineup alongside Gemini 4 Carbon and Gemini 4 Argon within a matter of days. Built for rapid execution across agentic workflows, Hydrogen brings Google’s latest model capabilities to everyday coding, software development and other tasks that require speed and efficiency. Unlike Carbon and Argon, which are designed for deep, long-horizon reasoning on complex problems, Hydrogen focuses on fast execution. It can write code, run tests, verify changes, and handle routine development tasks, either independently or alongside the other Gemini 4 models. Hydrogen also surpassed Argon on the Artificial Analysis Intelligence Index, scoring 51 against Argon’s 50 at the default high thinking level. Its score rises to 54 at the highest xhigh setting, according to Google. Faster coding and smarter execution Google designed Hydrogen to handle everyday developer workloads, including debugging, repository coding, web development and multi-step agentic tasks. It can also process text, images, audio, and video, with a combined input-output context window of up to 1 million tokens. The model offers four thinking levels: low, medium, high, and a new xhigh setting, allowing developers to adjust reasoning depth according to a task’s complexity, latency, and cost requirements. In Google’s evaluations, Hydrogen scored 57 percent on Terminal-Bench 4.1, more than three times the 18.2 percent recorded by Gemini 3.8 Flash. The company also says Hydrogen completes tasks 2.5 times faster than its predecessor at the default high setting. For broader knowledge work, the model can analyze documents, spreadsheets, charts, and recorded video to produce reports and structured outputs. Google also highlights improvements in tool use, coding accuracy, and its ability to recover when a tool call fails or a search returns no results. Orchestrating multiple models Google is also introducing Orchestrator in its Antigravity development platform, allowing developers to combine Gemini 4 Carbon and Hydrogen within a single session. Rather than relying on one model throughout a task, Orchestrator dynamically switches between the two based on the work required. Carbon can initially examine a codebase, resolve ambiguities, and create an execution plan, while Hydrogen takes over routine coding, package updates, and testing. If the faster model encounters difficulties, the task can return to Carbon for further reasoning and adjustments. On Terminal-Bench 4.1, Orchestrator achieved task success rates close to Carbon while consuming 24 percent less usage quota and completing tasks 19 percent faster on average, Google said. The approach is intended to reduce the need to use a high-reasoning model for every step of a development workflow, potentially lowering costs while maintaining performance. Google is rolling out Hydrogen across Google AI Studio, the Gemini API, Gemini for business, the Gemini app, and Google Search. It is also available to paid Google AI Pro and Ultra subscribers in Antigravity, while Orchestrator is initially rolling out to Google AI Ultra subscribers. The model’s API pricing is set at $1.50 per million input tokens and $7.50 per million output tokens across its 1-million-token context window. Get the latest in engineering, tech, space & science - delivered daily to your inbox.With over a decade-long career in journalism, Neetika Walter has worked with The Economic Times, ANI, and Hindustan Times, covering politics, business, technology, and the clean energy sector. Passionate about contemporary culture, books, poetry, and storytelling, she brings depth and insight to her writing. When she isn’t chasing stories, she’s likely lost in a book or enjoying the company of her dogs.