Skip to contentOceanToken

3 min read

Gemini 4 Hydrogen and Carbon: Google's Fast and Deep Models

Google shipped Gemini 4 Hydrogen on Nov. 12, two days after Gemini 4 Carbon went GA. What each model is for, what it costs, and how they work together.

By OceanToken Team

Colorful Google logo on the glass facade of Google's Park View office in Seattle

Google filled out its Gemini 4 lineup in one week. Gemini 4 Carbon, its deepest reasoning model, became generally available to paying developers on November 10. Two days later, on November 12, Google released Gemini 4 Hydrogen, the fast tier, across the Gemini API, the Gemini app and Search. With Gemini 4 Argon already GA since October, developers now have three Gemini 4 models to choose from.

The three Gemini 4 models

Model Role GA date API price per 1M tokens (in / out)
Gemini 4 Hydrogen Fast tier for everyday coding and agents Nov 12, 2026 $1.50 / $7.50, flat across a 1M-token context (reported)
Gemini 4 Argon Default choice for most software and knowledge work Oct 13, 2026 $2 / $10 introductory, then $4 / $20
Gemini 4 Carbon Deepest reasoning: long agentic coding, science, security Nov 10, 2026 $2 / $10 through Nov 30, then $4 / $20

Gemini 4 Hydrogen: the fast tier

Hydrogen (model ID gemini-4-hydrogen) replaces Gemini 3.8 Flash as Google's speed-first model. It is available on all Gemini API tiers, including the free tier, and is now the default model for the Antigravity agent in Gemini Managed Agents. It is also the default model for AI Mode in Google Search.

  • Four thinking levels: low, medium, high (the default) and a new xhigh. Each step up trades latency and tokens for more reasoning.
  • Speed: Google says Hydrogen finishes tasks 2.5 times faster than 3.8 Flash at its default level, and partners saw 2x to 2.8x faster runs with 12% to 23% fewer tokens.
  • Quality: 51 on the Artificial Analysis Intelligence Index at high, one point above Argon, and 54 at xhigh. On Terminal-Bench 4.1 it scores 57% against 18.2% for 3.8 Flash.

All benchmark numbers above are Google's own figures from the launch.

Gemini 4 Carbon: the deep tier

Carbon (model ID gemini-4-carbon) first shipped on October 28 to cyber defenders through Google's Fairwind Program. Since November 10 it is open to paying developers in Google AI Studio and Gemini for business, and to Google AI Ultra subscribers in the Gemini app, AI Mode and Antigravity. Google positions it for long-running agentic coding, multimodal knowledge work, scientific reasoning and cybersecurity defense.

Carbon spends more tokens reasoning than Argon, so Google still recommends Argon for most everyday engineering work and Carbon for the hardest tasks. Google reports first place for Carbon on CursorBench 4.0 and Zapier's AutomationBench, and lower attack success rates than Argon in its safety tests.

How Hydrogen and Carbon work together

The new Orchestrator mode in Google Antigravity runs Carbon as the planner and lets it hand routine steps to Hydrogen. Google reports that Orchestrator comes close to Carbon's own success rate on Terminal-Bench 4.0 while using less quota and finishing sooner. The same pattern works in your own code: route quick, well-defined steps to the fast model and keep the deep model for planning and hard calls.

Why it matters for developers

If you run Gemini 3.8 Flash today, Hydrogen is the natural upgrade: it is priced as a fast tier and Google claims large gains in both speed and agentic coding. Review Google's migration notes before switching, and check workloads on Gemini Managed Agents, where Hydrogen is now the Antigravity default. If you are testing Carbon, the introductory price ends November 30, after which it doubles to $4 / $20 per million tokens, so budget accordingly.

On OceanToken

We are adding Gemini 4 Hydrogen and Carbon to OceanToken. Once they are live they will appear on the Models page with their prices, and you will call them with the same API key and the same OpenAI-compatible API as every other model. Until then, Gemini 3.8 Flash runs with the same call; swap the model ID when the new ones are listed:

curl https://api.oceantoken.ai/v1/chat/completions \
  -H "Authorization: Bearer $OCEANTOKEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "google/gemini-3.8-flash", "messages": [{"role": "user", "content": "Hello"}]}'

Sources

  • News
  • New models
  • Google