Skip to content
Google

Gemini 4 Hydrogen

Google's fast workhorse and the successor to Gemini 3.8 Flash, under a new name: much stronger at agentic coding, at one flat price.

Quality

Modality

multimodal

Context

1M tokens

Access

closed

Fabian's Take

FM

"On raw model capability Google has been left behind by OpenAI and Anthropic, and that matters less than it sounds. A good harness and good connections to everything else you use are often worth more than the last few points of capability. I use a lot of Google services, so Hydrogen is an inexpensive, fast, capable coworker, especially inside the Antigravity harness. It's also excellent at watching and reviewing videos for me."

Gemini 4 Hydrogen came out on 12 November 2026. Google’s product names remain a puzzle, but this is the successor to Gemini 3.8 Flash: Google calls it a “major standalone upgrade” for teams moving from 3.8 Flash, and it takes Flash’s place in the Gemini app and in Search.

What changed

By Google’s own numbers it finishes tasks 2.5 times faster than 3.8 Flash at its default setting, scores 57% on Terminal-Bench 4.1 (a test of agents doing work on the command line) against 3.8 Flash’s 18.2%, and makes up facts less than a fifth as often. You can set how long it thinks before answering, in four steps from “low” to a new “xhigh”.

The price

$1.50 per million input tokens and $7.50 per million output, the same across the whole 1M-token context, with batch requests at half price. That’s twice 3.8 Flash’s introductory price and exactly what 3.8 Flash will cost from 1 January 2027. There’s a free tier on the API, and free users of the Gemini app get Hydrogen when the automatic model choice decides it fits the prompt.

Where it fits

On raw capability, Google has been left behind by the frontier models from OpenAI and Anthropic. That matters less than the benchmark tables imply. How clever the model is turns out to be only one part of a working setup. The other part is the software wrapped around it: what it can see, what it can reach, which of your other tools it can act on. In AI circles that wrapper gets called the harness, and for day-to-day work it’s often worth more than the last few points of raw capability.

That’s why this model works as a coworker for me. I use a lot of Google services, so it already sits next to the things I work on, and inside Antigravity, where it’s now the agent’s default for paid Google AI Pro and Ultra subscribers, it’s inexpensive, fast, and capable enough. It’s also excellent at watching and reviewing video, which is a job most models are still bad at.

Use it for volume work. For hard logic or high-stakes reasoning, step up to a frontier model and let Hydrogen handle everything around it.

The Verdict

Best for: High-volume execution work: coding, agents, copy, and anything where speed and price beat peak intelligence.

Pros

  • By Google's numbers 2.5 times faster than 3.8 Flash and far stronger at agentic coding (57% against 18.2% on Terminal-Bench 4.1)
  • One flat price across the whole 1M-token context, and a free API tier
  • Free users of the Gemini app get it when the automatic model choice picks it
  • Strong at watching and reviewing video
  • Already sits inside the Google services and tools you may be working in anyway

Cons

  • Twice 3.8 Flash's introductory price, at $1.50/$7.50
  • The new name doesn't tell you it's the successor to Flash
  • Output capped at 64K tokens unless you use long-decode mode

Specs

  • Pricing $1.50/M input, $7.50/M output, flat across the full context · batch and flex half price · cached input $0.075/M · free API tier
  • Cost Tier moderate
  • ⚡
    Speed Tier fast
  • License Proprietary
Developer Docs