Gemma 3 4B vs MiMo-V2.6-Pro

Pyyan / Compare / Gemma 3 4B vs MiMo-V2.6-Pro

2 of 5

Open-Weight Models · verified 29 Sept 2026

Gemma 3 4BGooglecurrent

XMiMo-V2.6-ProXiaomicurrent

Add another3 slots left

Specification	Gemma 3 4B	MiMo-V2.6-Pro
Summary	Multimodal and 140+ languages at 4B.	A trillion parameters under the MIT licence.
Context	—	1M
Input ($/Mtok)	—	Weights only
Output ($/Mtok)	—	Weights only
Licence	Gemma licence	MIT
Max output	—	Not published
Parameters	4B	1.02T total, 42B active
Released	—	21 Sep 2026
Model ID	—	XiaomiMiMo/MiMo-V2.6-Pro
Category	Open-Weight Models	Open-Weight Models
Official	Google ↗	Xiaomi ↗

Highlighted rows are where these differ.

Multimodal and 140+ languages at 4B.

Best for multilingual edge.

1.02T total parameters, 42B active per token, a 4.1% activation ratio
Text, image, video and audio in, with a 1M token context
Shipped with 7,000+ reinforcement learning environments and the framework that trained it

Best for multimodal work on your own cluster.