Google Rolls Out Gemini 4 Hydrogen to All Search Users, Saying It Finishes Tasks 2.5 Times Faster

Google on Nov. 12, 2026, released Gemini 4 Hydrogen for Search, the Gemini app and the API, saying the model finishes tasks 2.5 times faster than 3.8 Flash.

Share this article

Featured image: the patio at the Googleplex in Mountain View, California, in an August 2014 file photo. Photo: Jijithecat, Wikimedia Commons, CC BY-SA 4.0. Cropped to 16:9.

Google on Thursday, Nov. 12, 2026, released Gemini 4 Hydrogen, a new model it said is built for fast coding and other agent work. The company said Gemini 4 Hydrogen is now rolling out in Google Search for all users around the world.

Advertisement

The company priced the Gemini 4 Hydrogen application programming interface at $1.50 per million input tokens and $7.50 per million output tokens, a flat rate across a 1-million-token context window.

$1.50 / $7.50
Google’s published Gemini 4 Hydrogen API price, per million input and output tokens, flat across a 1-million-token context window
2.5x
How much faster Google said Hydrogen finishes tasks than Gemini 3.8 Flash, at Hydrogen’s default “high” thinking setting
57% vs. 18.2%
Terminal-Bench 4.1 scores Google reported for Hydrogen and for Gemini 3.8 Flash
Nov. 12, 2026
Date Google published the Hydrogen announcement; the company last updated the post on Nov. 13

All four figures are from Google’s blog post “Introducing Gemini 4 Hydrogen,” dated Nov. 12, 2026, and updated Nov. 13, 2026.

Where Google said Gemini 4 Hydrogen is available

In a blog post signed by Koray Kavukcuoglu, Google DeepMind’s senior vice president and Google’s chief AI architect, the company said Hydrogen “brings cutting-edge intelligence to everyone” and is intended for “everyday coding and agentic workflows.” Google contrasted that role with Gemini 4 Carbon and Gemini 4 Argon, which it said are meant for deeper, longer-running reasoning.

The company said Hydrogen was rolling out globally the same day across developer, enterprise and consumer products: the Gemini API in Google AI Studio and Android Studio; Gemini for business; the Gemini app; Google Search; Gemini in Google Sheets for eligible business subscriptions and for Google AI Pro and Ultra subscribers; and Google Antigravity for paid Pro and Ultra subscribers. In the Gemini app, Google said, free users would get access “via auto model selection when it’s the appropriate model for the prompt.” 9to5Google, reporting the same day, said the model had begun appearing in the Gemini app and that the in-app picker described Hydrogen as “Everyday help.”

A companion Google Search blog post on Nov. 12, by Robby Stein, a vice president of product for Search, called Hydrogen “the biggest model upgrade to Search yet” and said it was “available globally in Google Search for all users.” The Boston Report has not independently tested the rollout in every country or on every account type. The availability details above are what Google published.

The Hydrogen launch follows a narrower debut of Gemini 4 Carbon. The Boston Report first reported on Gemini 4 Carbon in late October, when Google described the model as a cybersecurity-focused system with limited access. On Nov. 10, Google said it was making Carbon available to paying developers and enterprises through the Gemini API, and to Google AI Ultra subscribers in Antigravity, the Gemini app and AI Mode in Search, two weeks after a release to cybersecurity defenders in its Fairwind Program. Gemini Omni 1.2 Flash, a separate Google model the company has described as a faster option for short video clips, remains in the same product family and is not replaced by Thursday’s announcement.

By the numbers: API prices and a coding score Google reported

Google priced Hydrogen at $1.50 per million input tokens and $7.50 per million output tokens, and it said that rate is flat across the model’s entire 1-million-token context window. The DeepMind model card for Gemini 4 Hydrogen, published Nov. 12, 2026, also describes a combined input-output context window of up to 1 million tokens, with a default of up to 64,000 synchronous output tokens and up to 1 million output tokens “via background long decode.”

Those Hydrogen prices can be compared with the Carbon rates Google published two days earlier. In the Nov. 10 Carbon post, Google said Carbon’s introductory API price — “the same introductory price as Gemini 4 Argon” — is $2 per million input tokens and $10 per million output tokens, and that after the introductory period expires on Nov. 30, 2026, the price “of $4 per 1M input tokens and $20 per 1M output tokens will apply,” for “the Carbon model intended for public access only.”

By The Boston Report’s calculation from those published figures, Hydrogen’s input and output prices are 25% lower than Carbon’s introductory rates and 62.5% lower than the Carbon rates scheduled to take effect after Nov. 30. In all three price pairs, the output rate is five times the input rate. Those percentages describe list prices only. They are not a finding about what any given customer will actually pay after discounts, committed-use deals or cached prompts, which Google’s posts do not itemize here.

Charts of Gemini 4 Hydrogen API prices and a coding-agent score: Google priced Hydrogen at $1.50 per million input tokens and $7.50 per million output tokens, compared with Gemini 4 Carbon’s $2 and $10 introductory rates and $4 and $20 rates after Nov. 30, 2026; Google also reported Terminal-Bench 4.1 scores of 57% for Hydrogen and 18.2% for Gemini 3.8 Flash
Sources: Google, “Introducing Gemini 4 Hydrogen,” Nov. 12, 2026, updated Nov. 13, 2026; and “Gemini 4 Carbon is now available,” Nov. 10, 2026, including the footnote on Carbon’s post-introductory price. The 25% and 62.5% price differences cited in the text are Boston Report calculations from those published rates. Chart: The Boston Report.

On Terminal-Bench 4.1, a coding-agent benchmark, Google said Hydrogen scored 57%, compared with 18.2% for Gemini 3.8 Flash, which the company called “more than tripling” the earlier result. A Boston Report calculation from those two figures is 3.13 times, which matches that description. Google also said Hydrogen, at its default “high” thinking level, “surpasses Argon on the Artificial Analysis Intelligence Index” with a score of 51 versus 50, and that it reaches 54 at a new “xhigh” thinking setting. Those benchmark numbers are Google’s. The Boston Report has not re-run the tests.

Orchestrator, speed claims and everyday Search

Google said it is also introducing Orchestrator in Google Antigravity, a setting that “dynamically switches between models like Carbon and Hydrogen within the same session.” On Terminal-Bench 4.0, the company said, Orchestrator “reached near-Carbon task success rates while, on average, consuming 24% less quota and completing 19% faster.” For work that uses several agents at once, Google said Carbon can coordinate while Hydrogen runs worker and verifier agents.

Advertisement

The company presented Hydrogen as a “major standalone upgrade” for teams moving from Gemini 3.8 Flash, saying it finishes tasks 2.5 times faster at its high default setting, “cuts fabricated claims by more than 5x,” and, in early testing by partners on “complex graphs, SQL, and data-heavy enterprise workflows,” ran “2x to 2.8x faster while reducing token usage by 12% to 23%” compared with 3.8 Flash. It also published a series of other leaderboard figures, including a No. 5 ranking on Agent Arena, class-leading or first-place scores on HLE-Diamond, the Vals Tax Agent Bench and the Vals Legal Research Bench, and 53.4% on a new Multimodal Knowledge Agent benchmark. All of those results are Google’s characterizations of third-party or partner evaluations. They are not independent audits by The Boston Report.

For consumers, Google said Hydrogen would power more visual Search responses and a “query fan-out” technique that it said would surface a wider range of links. 9to5Google reported that Hydrogen is the default model for AI Mode and is used for “your hardest questions” in AI Overviews. That additional product detail comes from 9to5Google’s account of the launch, not from a sentence in Google’s main Hydrogen post.

The release is part of a fast-moving stretch of business news about which companies can sell frontier models, at what price, and through which products. Readers can compare recent releases in our AI models guide and follow continuing coverage in our AI news center.

Safety figures Google cited

Google said Hydrogen uses the “latest Gemini 4 safety and security stack” and published a set of internal and benchmark comparisons with 3.8 Flash. Among them: the company said the model “reduces overreach in agentic tasks by up to 81% and under-service by up to 79%,” “cuts misaligned actions and task shortcutting across tool-use and coding evaluations by 12x,” is “up to 2x less likely to bypass permission boundaries,” and “achieves a 94% honesty rate on our internal agent honesty evaluation.” It also said Hydrogen is “more than 2.3x more resilient” on the Gray Swan indirect prompt-injection benchmark, and that updated safeguards kept “false refusals on HealthBench Pro under 0.4%.”

Those figures are Google’s own safety measurements. Several of them are described as “up to” improvements or as results on internal evaluations, which means the published post does not, by itself, show how the model performs in every real-world tool or in every country where Search is rolling out. What Thursday’s announcement does establish is the product Google says it has now put in front of a mass audience: a lower-priced, faster Gemini 4-class model in Search, the Gemini app and the paid API, two days after Carbon was opened more widely to paying developers.

Sources

Read next in Business & Finance

Get The Boston Report Briefing

Our free weekly email with the most important stories from Boston and the world. Unsubscribe anytime.

Free. We never sell your address. Privacy policy

Get The Boston Report Briefing
Our free weekly email with the most important stories from Boston and the world. Unsubscribe anytime.
Sign up