Beam 1.0 vs MiMo-V2.6-Pro

Pyyan / Compare / Beam 1.0 vs MiMo-V2.6-Pro

2 of 5

Open-Weight Models · verified 26 Oct 2026

RBeam 1.0Reflectioncurrent

XMiMo-V2.6-ProXiaomicurrent

Add another3 slots left

Specification	Beam 1.0	MiMo-V2.6-Pro
Summary	501B under Apache 2.0, with nothing asked in return.	A trillion parameters under the MIT licence.
Context	1M	1M
Input ($/Mtok)	$0.25	Weights only
Output ($/Mtok)	$1	Weights only
Licence	Apache 2.0	MIT
Max output	131K	Not published
Parameters	501B total, 23B active	1.02T total, 42B active
Released	16 Oct 2026	21 Sep 2026
Model ID	reflection/beam-1.0-501b-a23b	XiaomiMiMo/MiMo-V2.6-Pro
Category	Open-Weight Models	Open-Weight Models
Official	Reflection ↗	Xiaomi ↗

Highlighted rows are where these differ.

Apache 2.0, with no attribution clause and no revenue trigger, unlike Meta's Muse licence
52 layers, 256 experts, a 1M token context and a knowledge cutoff of 30 June 2026
Reflection claims comparable answers to leading open models with roughly three times fewer reasoning tokens
Artificial Analysis scores it 30 on the intelligence index, 19th of 120 models

Best for open weights with no licence conditions attached.

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.