Gemma 4 31B
Google's open-weight Gemma model, served free on several hosts.
Google, chat, open weights. Checked 30 October 2026.
Where to use it through an API
Hugging Facegoogle/gemma-4-31B-it
SambaNovagemma-4-31B-itFree key
The maker publishes its own benchmarks and limits on its model page.
How it compares
The models just above and below it on the Artificial Analysis Intelligence Index.
| Model | Index | Context | Speed |
|---|---|---|---|
| Gemini 4 Carbon | 59 | n/a | n/a |
| Gemini 4 Argon | 50 | 1M | 83/s |
| Gemini 3.8 Flash | 31 | 1M | 129/s |
| Gemini 3.5 Flash-Lite | 12 | 1M | 439/s |
Use Gemma 4 31B with every other AI at once
Put Gemma 4 31B in one panel and ChatGPT, Claude, Gemini or any of 300+ models in the others. One message goes to all of them, answers arrive side by side, and test mode times them and hides the names so you judge fairly. Pro is free with a follow on X, and you use your own keys.
Questions
Is there a free Gemma 4 31B API?
Yes, within limits: SambaNova give free keys that reach Gemma 4 31B.
What is the Gemma 4 31B model id?
Hugging Face: google/gemma-4-31B-it; SambaNova: gemma-4-31B-it.
How does Gemma 4 31B score?
The Artificial Analysis Intelligence Index had not scored it when we checked on 30 October 2026.
Can I use Gemma 4 31B with ChatGPT, Claude and Gemini at the same time?
Yes. MultiScreenAI puts Gemma 4 31B in one panel and other models in the others, sends one message to all of them and shows the answers side by side.