# Gemini 4 Hydrogen Gemini 4 Hydrogen is an AI model released by Google on Nov 12 2026. It has a 1M context window. Tracked results include 76.1% on AutomationBench (Artificial Analysis), 51.4% on Terminal-Bench-Science 0.1 and 77.6% on OSWorld 2.1 (offline). ## Facts | Field | Value | | --- | --- | | Model | Gemini 4 Hydrogen | | Developer | Google | | Release date | Thursday, Nov 12 2026 | | Licensing | Closed | | Context window | 1M | ## API pricing All rates in USD per 1,000,000 tokens, pay-as-you-go. | Tier | Input | Cached input | Output | | --- | --- | --- | --- | | Standard | $1.50 | $0.075 | $7.50 | Output prices include reasoning tokens. The pricing page lists a free tier at no charge. Free-tier content is used to improve Google's products; paid-tier content is not. Context caching is $0.075 per million tokens and applies to implicit caching only. Explicit caching is not supported. The page lists one paid text rate for the whole 1,048,576-token window, with no long-context multiplier. Verified November 12, 2026 against the first-party source: https://aistudio.google.com/docs/pricing#gemini-4-hydrogen ## Tracked benchmark scores | Benchmark | Score | Source | What it measures | | --- | --- | --- | --- | | Terminal-Bench-Science 0.1 | 51.4% | Lab | The same command-line setup as Terminal-Bench, pointed at scientific work: the AI has to drive research tooling and computational workflows through to a result, rather than administer a machine. Version 0.1 is the first release of the task set, and scores run lower than on the general board. Higher is better. | | OSWorld 2.1 (offline) | 77.6% | Lab | Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Version 2.1, in its offline setting. Scores are not comparable with version 2.0. Higher is better. | | AutomationBench (Artificial Analysis) | 76.1% | Lab | The same business-workflow test, run by Artificial Analysis rather than by the lab, and graded strict pass@1: a workflow counts only if the whole of it is right on the first attempt. Read it as a separate board from the lab-run AutomationBench scores. Higher is better. | | Vending-Bench 2 (USD) | 14549 | Lab | ending bank balance in US dollars in Andon Labs’ vending-business simulation; higher is better | | Chartography (no tools) | 66.6% | Lab | The same chart-centred test with no tools: the AI has to read each chart unaided. Scores run far lower than the with-tools version, so read the two as separate tests. Higher is better. | ## About Gemini 4 Hydrogen Gemini 4 Hydrogen was announced on November 12, 2026 as Google's Gemini 4 model for everyday coding and agentic workflows. The model card describes it as adapted from Gemini 4 Carbon. Google presented Carbon and Argon as the frontier models for long-horizon reasoning, and Hydrogen as the faster model for writing code, running tests and multi-step tasks. The announcement rolled it out the same day to Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Search, and to paid Google AI Pro and Ultra subscribers in Google Antigravity. Google's model card gives Hydrogen a 1,048,576-token window shared by input and output, with synchronous replies capped at 65,536 tokens and a background long decode up to the full window. Four thinking levels — low, medium, high and xhigh — set the reasoning depth, and high is the default. Google said that default finished tasks 2.5 times faster than Gemini 3.8 Flash. Its launch evaluation reported 77.6% partial credit on the offline subset of OSWorld 2.1, 66.6% on Chartography without tools, and an ending bank balance of $14,549 on Vending-Bench 2. Google also said Hydrogen scored 51 on the Artificial Analysis Intelligence Index at the default high setting, compared with 50 for Argon, and 54 at xhigh. Those index figures are the lab's claim. On the Gemini API pricing page the paid rates were $1.50 per million input tokens and $7.50 per million output tokens, including thinking, with implicit cache reads at $0.075, and a free tier was listed beside them. ## Questions and answers ### When was Gemini 4 Hydrogen released? Gemini 4 Hydrogen was released by Google on Thursday, Nov 12 2026. ### Who made Gemini 4 Hydrogen? Gemini 4 Hydrogen was built by Google. Builds the Gemini family of models through Google DeepMind. Integrates AI across Google products. ### How much does Gemini 4 Hydrogen cost? Gemini 4 Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens through the Google API. Cached input is $0.075 per million tokens. Output prices include reasoning tokens. Rates are pay-as-you-go API prices verified against Google's published pricing on November 12, 2026. ### What benchmark scores did Gemini 4 Hydrogen get? Gemini 4 Hydrogen reports 5 tracked benchmark scores — Terminal-Bench-Science 0.1: 51.4%; OSWorld 2.1 (offline): 77.6%; AutomationBench (Artificial Analysis): 76.1%; Vending-Bench 2 (USD): 14549; Chartography (no tools): 66.6%. Tracked scores may come from lab reports or independent benchmarks; source details accompany the benchmark data. It holds the best score among all models tracked here on AutomationBench (Artificial Analysis). ### What is the context window of Gemini 4 Hydrogen? Gemini 4 Hydrogen has a context window of 1M. Providers may specify separate input and output token limits. ### Is Gemini 4 Hydrogen open source? No. Gemini 4 Hydrogen is a closed model. The weights are not published — it is available only through the provider's own API, apps, or partner platforms. ### What came before and after Gemini 4 Hydrogen? Google's previous tracked release was Gemini 4 Carbon on Oct 28 2026, 15 days earlier. It is the most recent Google model tracked on AI Release Tracker. --- Canonical page: https://aireleasetracker.com/model/google/gemini-4-hydrogen Site index: https://aireleasetracker.com/llms.txt Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.