Gemini 4 Hydrogen, Carbon and Argon Are Now Available
All three Gemini 4 models are in the AI Data Stream model picker. Hydrogen is $1.50 / $7.50 per million tokens; Carbon and Argon are $2 / $10 until 30 November, then double.
Google has released three Gemini 4 models, and all three are now in the AI Data Stream model picker: Gemini 4 Hydrogen, the fast and efficient one, Gemini 4 Carbon, which Google calls its strongest reasoning model, and Gemini 4 Argon, which costs the same as Carbon but answers faster.
What Changed for You
Nothing you need to do. The three new models are extra choices in the picker.
- Gemini 3.8 Flash stays the default for Google. If you have already chosen a default model, it stays as it is.
- Gemini 3.1 Pro stays available, now labelled as the previous generation, so you can compare it with Carbon and Argon before moving over.
- Gemini 2.5 Pro has been retired. Google deprecated it on 2 November, and from 1 December 2026 answers its requests with Gemini 3.8 Flash instead. If it was your default model, Gemini 4 Hydrogen is now used automatically.
Your API key, settings and conversations all remain the same.
Pricing
Gemini 4 Hydrogen (per million tokens):
- Input: $1.50
- Cached input: $0.075
- Output: $7.50
Gemini 4 Carbon and Gemini 4 Argon (per million tokens), at an introductory price through 30 November 2026:
- Input: $2, then $4 from 1 December 2026
- Cached input: $0.10, then $0.20
- Output: $10, then $20
How they compare:
- Carbon and Argon vs Gemini 3.1 Pro: until 30 November they are slightly cheaper than 3.1 Pro ($2 / $12 for prompts up to 200k tokens). From 1 December they cost more, at $4 / $20.
- Hydrogen vs Gemini 3.8 Flash: Hydrogen costs twice as much as 3.8 Flash today ($0.75 / $3.75). From 1 January 2027, when 3.8 Flash’s introductory price ends, the two cost the same for input and output. That is why 3.8 Flash remains the default for now.
- Hydrogen vs Gemini 2.5 Pro: if you are migrated from 2.5 Pro ($1.25 / $10 for prompts up to 200k tokens), input costs slightly more and output costs a quarter less.
All three have a context window of one million tokens, and the price is the same across the whole window. There is no higher rate for long prompts, as there is on Gemini 3.1 Pro.
Two things to know before you switch:
- Carbon and Argon need a paid Google API key. They are not on Google’s free tier. Hydrogen is.
- Thinking is billed as output. All three models reason before they answer, and Google bills those thinking tokens at the output rate. Carbon spends more thinking tokens per answer than Argon, so at the same price per token it costs more per question and takes longer to reply.
Caching is automatic: Google applies the cached rate to the repeated part of prompts of 4,096 tokens or more. Since every follow-up question re-reads the conversation so far, long analytics conversations benefit without anything to switch on.
As always, AI Data Stream is bring-your-own-key: you are billed by Google directly at their published rates.
What’s New
How Google describes the three models:
- Gemini 4 Hydrogen is built for speed and efficiency in everyday work with tools. Google says it cuts fabricated claims by more than 5x compared with Gemini 3.8 Flash by flagging missing context instead of guessing (it gives no benchmark for that figure), and recommends it as the upgrade from 3.8 Flash and from earlier Gemini 3 and 2.5 models.
- Gemini 4 Carbon is Google’s strongest reasoning model, built for sustained agentic work. It spends more thinking tokens exploring hypotheses, verifying its work and checking edge cases before it responds, so each turn uses more tokens and takes longer than on Argon.
- Gemini 4 Argon is Google’s efficient Gemini 4 reasoning model, which it calls ideal for most advanced software engineering and knowledge work. It responds faster and spends fewer tokens per turn than Carbon.
That matters for AI Data Stream because every analysis is a chain of tool calls against your connected data sources. A model that plans those calls well, and says so when data is missing instead of guessing, gives better answers.
Google Search grounding works with all three, as it does across our Gemini lineup.
See Google’s documentation for Gemini 4 Hydrogen, Gemini 4 Carbon and Gemini 4 Argon for the full details.
Which One to Pick
- Gemini 3.8 Flash for the default: the lowest price of the main models until the end of the year.
- Gemini 4 Hydrogen for everyday questions and reports where you want the newer generation. It is worth trying on your own data before assuming you need Carbon or Argon.
- Gemini 4 Argon for deeper, multi-step analysis.
- Gemini 4 Carbon for the hardest analysis, when you can wait longer for the answer.
- Gemini 3.5 Flash Lite or 2.5 Flash Lite for the fastest, cheapest answers.
The Full Gemini Lineup
- Gemini 3.8 Flash - Best balance (default)
- Gemini 4 Hydrogen - Fast, efficient
- Gemini 4 Carbon - Most advanced
- Gemini 4 Argon - Advanced
- Gemini 3.1 Pro - Previous generation
- Gemini 3.5 Flash Lite - Fastest
- Gemini 2.5 Flash Lite - Cheapest
Switching Models
- Go to Settings > AI Providers
- Click your Google provider
- Select your preferred model from the dropdown
- Save
Or simply select it from the model picker when starting a new conversation.
Questions?
If you have any questions, contact us or check our AI providers documentation.
Related Posts
Gemini 3.8 Flash Is Now Available
Google released Gemini 3.8 Flash and it is now the recommended Google model in AI Data Stream, at the same introductory price as 3.7 Flash. 3.7 Flash stays available, 2.5 Flash is retired with automatic migration.
Gemini 3.7 Flash Is Now Available
Google released Gemini 3.7 Flash, and we have refreshed the whole Gemini lineup in AI Data Stream: 3.7 Flash and 3.1 Pro replace the old previews, with automatic migration and no action needed on your part.
Claude Fable 5.2 Is Now Available
Fable 5.2 is in the model picker at $8/$40 per million tokens, 20% less than 5.1, with big gains in reading charts. Fable 5.1 and 5 switch to it automatically.