Third-party API service says the model is live
AIHubMix, a third-party API service, said on its official X account at 1:37:55 a.m. Nov. 13 Japan time (Nov. 12) that "Gemini 4 Hydrogen," a model described as an unannounced Google model, is now available through its API. AIHubMix offers AI models from many companies through a single access point. Its X profile says users can access more than 800 models, including GPT, Claude, Gemini and DeepSeek, with one key and pay-as-you-go billing.
The post describes the model as "Google’s new model" that "is built for tough coding problems and long running engineering tasks." That is AIHubMix's description, and no Google material with the same description could be found. The post does not give a model ID, pricing or terms of use, and it does not link to any Google announcement or document. ORIENS has not tested whether the model can actually be called through AIHubMix's API.
Gemini 4 Hydrogen is live on AIHubMix Google’s new model is built for tough coding problems and long running engineering tasks Try it through our API. What’s your first prompt gonna be? https://t.co/cg3NYAMDcp
Leaker says it "looks like" the model was added to Vertex
About eight minutes later, at 1:45:58 a.m. Nov. 13 Japan time (Nov. 12), the leaker @notjazii, who posts information about unreleased models, quoted the AIHubMix post and wrote "Gemini 4 hydrogen is live." The leaker said the model appeared to have been added to Vertex AI, Google Cloud's development platform, and that AIHubMix had already integrated it into its service. That is the leaker's own impression, phrased as "looks like," and the post did not point to any Google material showing the addition.
On speed, the post only says tests were still running and that the model was "super fast." The leaker said results would be posted once the tests were done, and the post contains no figures on speed.
Gemini 4 hydrogen is live looks like model was added on google vertex and aihubmix already integrated it into their platform running some tests right now and man it's super fast will post results as soon as it's done cooking https://t.co/XbpopkJHQz https://t.co/OO2587ugQ6
Name is absent from Google's official model lists
Google has not announced "Gemini 4 Hydrogen." As of early Nov. 13 Japan time, the Gemini API model list for developers showed only two models in the Gemini 4 family, "Gemini 4 Carbon" and "Gemini 4 Argon," and did not include Hydrogen. The changelog at the same time ran through Nov. 11 and recorded that Gemini 4 Carbon became generally available on Nov. 10, but it made no mention of Hydrogen.
The URL of the Vertex AI model list now leads to the model list page of Google Cloud's "Gemini Enterprise Agent Platform." That page also covers only Carbon and Argon in the Gemini 4 family, and the Hydrogen name does not appear. The leaker's claim of an addition to Vertex cannot be confirmed from the official list. However, the list covers released models, so it cannot show whether the model is being offered early to a limited set of users.
Posts in early November called it a "Gemini 4 Flash" code name
The Hydrogen name had appeared in posts on X since late October. On Nov. 5, ORIENS reported on posts saying Hydrogen was the code name for "Gemini 4 Flash" and was being A/B tested in the developer environment "Google AI Studio" and on the comparison site "Arena." None of those posts offered evidence.
The AIHubMix post refers to "Gemini 4 Hydrogen" and does not use the word "Flash." There is no official information on whether Hydrogen corresponds to the lightweight Flash model or is a different model. Until Google makes a formal announcement, the name, where it is offered, pricing and performance all remain unconfirmed.
Addendum: On Nov. 13, Google officially released the model (Google: Introducing Gemini 4 Hydrogen).
- Post says "Hydrogen" is the code name for Gemini 4 FlashRumor
- Posts say Hydrogen is in A/B testing on AI Studio and ArenaLeak
- Google makes Gemini 4 Carbon generally availableOfficial
- AIHubMix posts that "Gemini 4 Hydrogen" is availableThird party
- @notjazii posts about a Vertex AI addition and speedLeak






