prismaquant
Mixed-precision quantization for LLMs. Every layer refracts into a different format based on its sensitivity. Native compressed-tensors export, validated on Qwen3.6-35B-A3B MoE with MTP speculative decoding.
View on Github
Activity
- Latest release
- 1d ago
- Total releases
- 6
- Cadence
- ~daily
- Last 12 months
- 6
Reach
- Stars
- 98
Details
- License
- Apache-2.0
- First release
- Jul 30, 2026
Releases
| Version | Released | |
|---|---|---|
0.4.1
patch
| ||
0.4.0
minor
| ||
0.3.1
patch
| ||
0.3.0
minor
| ||
0.2.1
patch
| ||
0.2.0
initial
|