v1.10.0 prebuilt-wheels-cu13
high risk
unslothai/unslothSep 26, 2026 → Sep 27, 20261 day between releases
What's New
Upgraded to CUDA 13 wheels with flash-attn 2.8.4, causal-conv1d 1.7.0, and mamba-ssm 2.3.2.post1 for PyTorch 2.13/2.14 on Python 3.13
Breaking Changes
- Requires Python 3.13
- Requires PyTorch 2.13 or 2.14
- Upgraded flash-attn to 2.8.4
Improvements
- Prebuilt wheels for faster installation on Linux x86_64 with CUDA 13 support
- Optimized for PyTorch 2.13 and 2.14
Upgrade Notes
Ensure your environment supports Python 3.13 and CUDA 13. This release targets PyTorch 2.13 and 2.14.
The Two Releases
v1.10.0
v1.10.0
Sep 26, 2026· 26 days ago
A performance-focused release introducing progressive analytics loading, media optimization, API idempotency, and significant improvements to realtime messaging capabilities.
Full release notesprebuilt-wheels-cu13
Flash-Attention2, Causal-Conv1D, Mamba_SSM Binaries
Sep 27, 2026· 25 days ago
Minimal release notes: "Prebuilt Linux x86_64 CUDA 13 wheels for flash-attn 2.8.4, causal-conv1d 1.7.0 and mamba-ssm 2.3.2.post1, built for PyTo"
Full release notes