Running
1
SWAN: Data-Free Mixed-Precision Quantization
🦢
Quantize large language models without data using mixed‑precision
AI & ML interests
Model Quantization
Recent Activity
View all activity
Organization Card
baa.ai
Smaller. Smarter.Sovereign.
Making frontier models run anywhere
We build open tools for efficient AI deployment. Our research focuses on quantization methods that preserve model quality while dramatically reducing hardware requirements — bringing 400B+ parameter models to a single machine.
baa.ai · SWAN Paper · GitHub
models 11
baa-ai/Llama-3.1-70B-Instruct-SWAN-5bit
Text Generation • 71B • Updated
baa-ai/Llama-3.3-70B-Instruct-SWAN-5bit
Text Generation • 71B • Updated
baa-ai/Llama-4-Scout-17B-16E-Instruct-SWAN-4bit
18B • Updated
baa-ai/Llama-4-Maverick-17B-128E-Instruct-SWAN
Text Generation • 57B • Updated
• 40
baa-ai/Llama-4-Maverick-17B-128E-Instruct-SWAN-4bit
Text Generation • 51B • Updated
• 58
baa-ai/MiniMax-M2.5-SWAN-4bit
Text Generation • 229B • Updated
• 168
baa-ai/GLM-4.7-Flash-SWAN-4bit
Text Generation • 30B • Updated
• 173 • 1
baa-ai/Qwen3-30B-A3B-SWAN-4bit
Text Generation • 31B • Updated
• 30 • 1
baa-ai/Qwen3-8B-SWAN-6bit
Text Generation • 8B • Updated
• 20 • 1
baa-ai/Qwen3.5-397B-A17B-SWAN-3bit
Text Generation • 396B • Updated
• 858 • 1
datasets 0
None public yet