DeepSeek V4 Flash Vision Exp vs MiMo-V2.6-Pro

Pyyan / Compare / DeepSeek V4 Flash Vision Exp vs MiMo-V2.6-Pro

2 of 5

Open-Weight Models · verified 29 Sept 2026

DeepSeek V4 Flash Vision ExpDeepSeekdeprecated

XMiMo-V2.6-ProXiaomicurrent

Add another3 slots left

Specification	DeepSeek V4 Flash Vision Exp	MiMo-V2.6-Pro
Summary	Experimental, says so in the name, and shipped anyway.	A trillion parameters under the MIT licence.
Context	128K	1M
Input ($/Mtok)	$0.44	Weights only
Output ($/Mtok)	$1.32	Weights only
Licence	Open weights	MIT
Max output	32K	Not published
Parameters	Not published	1.02T total, 42B active
Released	21 Aug 2026	21 Sep 2026
Model ID	deepseek-ai/DeepSeek-V4-Flash-Vision-Exp	XiaomiMiMo/MiMo-V2.6-Pro
Category	Open-Weight Models	Open-Weight Models
Official	—	Xiaomi ↗

Highlighted rows are where these differ.

Best for cheap multimodal trials.

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.