Gemini 4 Hydrogen limits and pricing

On the paid Gemini API, Gemini 4 Hydrogen costs $1.50 per million input tokens, $0.075 for cached input and $7.50 for output, including thinking tokens, per Google's Gemini Developer API pricing page and the Hydrogen what's-new guide. That price is flat across the whole 1 million-token window. There is no launch discount and no long-prompt surcharge. A medium month at those list rates comes to $67.50, our math.

Cached input is implicit only. Google says explicit caching is not supported, and implicit caching is on by default for prompts of at least 4,096 tokens, at 95% off the input price (that is the $0.075 figure). The API free tier lists Hydrogen as free of charge and rate-limited. Batch and Flex tiers are half the standard price; Priority is higher. We list the standard paid tier, which is what our Hydrogen model page uses.

Who gets Gemini 4 Hydrogen in the Gemini app

This table follows Google's subscriptions page and its Gemini Apps help article on models, both read Nov 12, 2026. Google's help article says free users (no Google AI plan) get Auto or Flash-Lite. When Auto is on, most prompts use Flash-Lite, and prompts that need more reasoning may use Hydrogen. If Auto is off, every response uses Flash-Lite. Plus, Pro and Ultra can pick Hydrogen in the model dropdown.

Usage in the Gemini app is still compute-based. Google's subscriptions footnote says limits factor in prompt complexity, the features you use and chat length, and that they refresh every 5 hours until you hit a weekly limit. Google does not publish a message count. Plus is 2x Free, Pro is 4x Free, and Ultra is 5x or 20x Pro, per the same page. If a paid plan runs out, chats continue on Flash-Lite, per the help article.

Google Antigravity and Search

Paid Google AI Pro and Ultra subscribers can use Hydrogen in Google Antigravity, per Google's Hydrogen announcement and the Antigravity blog. The Antigravity blog also says Orchestrator (Argon) is rolling out for paid Pro and Ultra, and Orchestrator (Carbon) for Ultra only. We have not independently timed those flows.

Hydrogen is in Google Search for all users, Google says, and in Gemini in Google Sheets starting with Pro and Ultra. Search limits are separate from the Gemini app caps above.

What changed from Gemini 3.8 Flash

Google calls Hydrogen the recommended upgrade from Gemini 3.8 Flash. Two things changed on price and one on the window.

• List price: Flash is $0.75 input and $3.75 output through Dec 31, 2026, then $1.50 and $7.50 from Jan 1, 2027, per Google's pricing page. Hydrogen is already at $1.50 and $7.50, so it costs more than Flash until that promo ends, and the same after.
• No intro promo: Hydrogen has no lower launch rate. The $1.50 / $7.50 figure is the ongoing list price.
• One shared 1M window: Google's what's-new guide says Flash split 1M input and 64K output, while Hydrogen shares one 1,048,576-token window across input, thinking and output, with long decode up to that cap (65,536 tokens in a single synchronous response; longer outputs use background execution).

Google reports Hydrogen scored 57% on Terminal-Bench 4.1, against 18.2% for Gemini 3.8 Flash, and finished tasks 2.5 times faster at its high default setting. Those are Google's figures, from the Nov 12 announcement. We do not yet have LMArena ranks for Hydrogen, so we are not naming a quality winner from our usual boards. For a side-by-side of the list prices, see Hydrogen vs Gemini 3.8 Flash.

Thinking levels and the context window

The API has four thinking levels: low, medium, high (the default) and a new xhigh tier, per the Hydrogen what's-new guide. Higher levels spend more thinking tokens, and those tokens are billed at the output price. Google's advice is to start at high and reserve xhigh for hard, multi-step work, because xhigh raises latency and token use. The Gemini app help article lists Low, Medium and High only, and says a manual switch with no level picked defaults to Low. That is the app, not the API.

Hydrogen vs Argon, Carbon, Sonnet and Terra

Each of these picks is about official list price and who can use the model. We are not inventing quality scores.

• Vs Gemini 3.8 Flash: Hydrogen is the one to use for new work. Flash is cheaper through Dec 31, 2026, then the two APIs cost the same. Winner for new work: Hydrogen. Winner on price until Dec 31: Flash. Full Hydrogen vs Flash comparison.
• Vs Gemini 4 Argon: Hydrogen wins on price ($1.50/$7.50 against Argon's launch $2/$10, which doubles on Dec 1) and on who can use it (Free and Plus included). Argon is Google's efficient frontier model for harder long-horizon work, on Pro and Ultra only. Winner for everyday work and a lower API bill: Hydrogen. Hydrogen vs Argon.
• Vs Gemini 4 Carbon: Carbon is Google's strongest Gemini 4 model. It joined the paid API on Nov 10, 2026, at the same launch price as Argon, per Google's Carbon availability post and the pricing page. In the Gemini app it is Ultra only. On the LMArena leaderboard dataset (CC BY 4.0), among models we price, Carbon is #1 of 54 for chat, #1 of 54 for roleplay, #3 of 54 for writing, #3 of 40 for agents and #4 of 53 for coding. Argon is #1 of 54 for writing, #1 of 52 for reasoning and #2 of 54 for chat. Hydrogen is not ranked yet. Winner for everyday work and a lower bill: Hydrogen. Winner at the same launch price for quality: Carbon, which leads Argon on more of the shared LMArena tasks. Hydrogen vs Carbon and Carbon vs Argon.
• Vs Claude Sonnet 5.5: Hydrogen costs less at list price ($1.50/$7.50 against $2/$10). Sonnet 5.5 already has LMArena ranks; Hydrogen does not, so we are not naming a quality winner. Winner on price: Hydrogen. Hydrogen vs Sonnet 5.5.
• Vs GPT-6.1 Terra: Terra is the closest-priced OpenAI model in our table. It wins on short-prompt price ($1/$5 against $1.50/$7.50). Hydrogen wins on long prompts: it stays $1.50/$7.50 across the full 1M window, while Terra jumps to $2 input and $7.50 output above 272K, per OpenAI's pricing page. Our pick: Terra for typical short-context bills, Hydrogen when you regularly send long context. Hydrogen vs GPT-6.1 Terra.

Carbon also now has a head-to-head against Claude Fable 5.2, which is the pair we held back until Carbon was on Google's official pricing page: Carbon vs Claude Fable 5.2.