Contact Corrections especially welcome

Contact Us

This is an independent reference, not Google. If you've found an error, have a suggestion, or want to reach us about the site itself, this is the place — but if you need help with Gemma 4 itself, the official channels below will get you a better answer faster.

🧭

Looking for Gemma support? You probably want Google, not us

We don't build Gemma and can't fix bugs, restore accounts, adjust rate limits or change model behaviour. These are the channels that can.

If you need…Go toWhy there
Official documentation ai.google.dev/gemma Authoritative and current; this site summarises it
A problem with the weights Hugging Face discussions Google and the community respond on the model repositories
API keys, quotas or billing Google AI Studio Only Google can see or change your project
A runtime bug The runtime's own tracker — Ollama, vLLM, llama.cpp, LM Studio Most "Gemma problems" are runtime configuration
Licensing certainty The licence text, and a lawyer Our summaries are careful but they aren't legal advice
To try the obvious fixes first Troubleshooting · FAQ Most issues are context length, GPU offload or a stale transformers version
Reach us

Send a message

Corrections are the most useful thing you can send. If a number here disagrees with a source you trust, we want to know.

0 characters

Or email REPLACE_WITH_YOUR_EMAIL directly.

Most useful

Reporting a correction

This site carries a lot of numbers, and some of them will go stale or turn out wrong. A good correction report includes:

  • ✅ Which page and which figure
  • ✅ What it should say instead
  • ✅ A link to the source you're citing
  • ✅ The date, if the source is time-sensitive

Benchmark figures, pricing and model rankings all decay. Sourced corrections get actioned fastest.

Can't help with

Out of scope

  • ⚠️ Google account, API key or billing problems
  • ⚠️ Bugs in Gemma itself, or in Ollama, vLLM or llama.cpp
  • ⚠️ Legal advice on licensing
  • ⚠️ Debugging your specific deployment
  • ⚠️ Anything requiring access to your project or hardware
What happens next

Setting expectations honestly

Corrections

Checked against the source you cite. If it holds up, the page is updated and the change noted. This is the category most likely to get a response.

💭

Suggestions

Read and considered, though not everything gets built. Requests for a topic the site doesn't cover are genuinely useful signal.

🚫

Support requests

Redirected to the channels above, which will help you faster. We can't see your API project, your hardware or your logs.

🔐
On what you send. Please don't include API keys, credentials, or anything confidential — a message form is the wrong place for any of them. If you're reporting a problem that involves a key, the right first step is to rotate it, not to send it.
ℹ️
Not affiliated with Google. This is an independent community reference. It isn't operated, endorsed or reviewed by Google or Google DeepMind, and messages sent here don't reach them. Gemma and Google are trademarks of Google LLC. For anything official, start at ai.google.dev/gemma. More on what this site is on the about page.

Most questions are already answered.

134 of them, searchable — and the troubleshooting tables cover the common failures.