state-spaces / mamba3-siso-893m
Other2842Updated Jul 28, 2026
Downloads
284
Likes
2
Last updated
Jul 28, 2026
License
apache-2.0
mamba3-siso-893m is a pretrained causal language model with 893M parameters. It is built from stacked blocks, each containing a Mamba-3 SISO mixer followed by a gated MLP. It contains no attention layers and is released with BF16 weights in the public mambassm checkpoint format.
Tasks & library
- mamba-ssm
Tags
- mamba-ssm
- pytorch
- mamba
- mamba-3
- state-space-model
- causal-lm
Technical specifications
ParametersNot available
Librarymamba-ssm
LanguagesNot available
Licenseapache-2.0
