X Debate: Unannounced Gemini 4 Hydrogen Shows Up on Google Cloud's Official SKU List at $1.50 Input and $7.50 Output per Million Tokens, 62.5% Below Gemini 4 Carbon's December Price, as Trackers Say It Already Answers on Vertex AI

Background and Pain Points

Google has moved fast this week, launching Gemini 4 Carbon around November 10 and Gemini Decision 1 on November 11 (US time). Developers have been waiting for a fast, cheap Gemini 4 tier, because coding, agent loops and batch jobs care most about unit price and latency, and Carbon and Argon rise to $4 input and $20 output per million tokens on December 1. Around midday on November 12 (Singapore time), Logan Kilpatrick posted the single word "Gemini" (about 475K views), and @aitrackerbot then said it had called gemini-4-hydrogen on Vertex AI, setting off a wave of reposts and speculation in English and Chinese.

What Is Actually Known

Only the billing entries can be verified. Google Cloud's public SKU page lists Hydrogen input, cached input and output SKUs under the Gemini API service, covering text, image, audio and video input across Standard, Batch, Flex, Priority and Contributor tiers, all in the global region. That shows billing is set up, but a billing entry is not a launch: Google has not published a context length, modality details, rate limits or availability. The claims that the ID was promoted from an -eap suffix and that the model calls itself Gemini 4 Hydrogen come from third-party screenshots that we could not reproduce, and Vertex access may be limited to some projects.

Benchmarks and Price Comparison

Google has published no benchmarks. Only prices can be compared, per Google: Hydrogen at $1.50 input and $7.50 output; Gemini 4 Carbon and Argon at $2/$10 now and $4/$20 from December 1; Gemini 3.8 Flash at $0.75/$3.75 this year and $1.50/$7.50 from January 1, 2027. For 1 million input and 200K output tokens, Hydrogen comes to about $3.00, against about $4.00 for Carbon now and $8.00 from December. Some replies argue that Hydrogen simply matches 3.8 Flash's list price, so the Flash tier is not getting cheaper; others say it is still a good deal if quality is close to Carbon. Speed and coding results circulating in the thread are single personal tests, not benchmarks.

What Developers Can Do Now

Do not hard-code gemini-4-hydrogen in production yet: it is not on the Gemini API models list, and its name and price may change. Reserve a fast-tier slot in your model router so a launch only needs a config change, measure per-task cost of Carbon, Argon and 3.8 Flash on your own eval set (Google notes that output prices include thinking tokens), and watch the ai.google.dev models and pricing pages for an official announcement before migrating.