Question 1

What is Gemma 4 31B good for?

Accepted Answer

The largest dense model of the open-weight Gemma 4 family (31 billion parameters) – server-grade performance that can also run on your own hardware. 4 of 5 outputs load without errors; 1 more ran after one fix by the same model. Its best technical result was on the “New York & Central Park map” prompt (98/100). Did not work (original output): Chess with an engine.

Question 2

How did Gemma 4 31B do in the AI Lab tests?

Accepted Answer

4 of 5 outputs load without errors; 1 more ran after one fix by the same model. Its best technical result was on the “New York & Central Park map” prompt (98/100). Did not work (original output): Chess with an engine. In the objective perft chess test it passed 0 of 6 positions. Average time per prompt 4 min 55 s – rank 16 of 27 models by speed. Average cost per prompt $0.00 – rank 1 of 27 models from the cheapest. Overall score 47/100 – rank 20 of 27 models. Technical health (speed, errors, accessibility, mobile) 89/100 – rank 16 of 27 models. Automatic UX rating 76/100 – rank 16 of 27 models.

Question 3

How much does Gemma 4 31B cost?

Accepted Answer

Through the API Gemma 4 31B costs $0 per million input and $0 per million output tokens (price list as of 17/10/2026).

Question 4

When was Gemma 4 31B released?

Accepted Answer

Gemma 4 31B by Google was released on 02/04/2026.

Gemma 4 31B

What is Gemma 4 31B good for?

How did Gemma 4 31B do in the AI Lab tests?

Outputs

Solar System

GTA-style game

Monitor test

New York & Central Park map

Chess with an engine