Zyphra Releases ZUNA1.1: An Apache 2.0 EEG Foundation Model With Variable-Length Inputs From 0.5 To 30 Seconds
Zyphra's ZUNA1.1 just 6x'd its input window. Here's why variable-length EEG models matter for brain-computer interfaces.
Why it matters
Zyphra expanded its open-source EEG foundation model to accept inputs from 0.5–30 seconds (vs. fixed 5s), improving flexibility for medical AI without sacrificing performance. This signals growing specialization in biomedical model infrastructure.
The key facts
12 to knowZUNA1.1 released July 16, 2026 under Apache 2.0
380M parameter masked diffusion autoencoder
Variable-length input support: 0.5–30 seconds (6x range vs. ZUNA1's fixed 5s)
Supports arbitrary scalp-EEG channel layouts
NMSE performance held or improved despite expanded input range
Open-source biomedical foundation model for EEG reconstruction, denoising, upsampling
ZUNA1.1 released July 16, 2026 under Apache 2.0 license
380M parameter masked diffusion autoencoder architecture
Variable-length input support: 0.5–30 seconds (vs. ZUNA1's fixed 5-second constraint)
Capabilities: EEG reconstruction, denoising, upsampling across arbitrary channel layouts
Performance: NMSE maintained or improved despite input range expansion
Open-source positioning competes with closed biomedical AI models
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Zyphra released ZUNA1.1 on July 16, 2026, under the Apache 2.0 license. The 380M masked diffusion autoencoder reconstructs, denoises, and upsamples scalp-EEG across arbitrary channel layouts. It accepts variable-length inputs from 0.5 to 30 seconds, against ZUNA1's fixed five seconds. Reported NMSE…