GLM
Collection
Baa.ai quantized versions of GLM models • 4 items • Updated
GLM-4.7-Flash — 19GB (MLX)
Mixed-precision quantized version of zai-org/GLM-4.7-Flash optimised by baa.ai using a proprietary Black Sheep AI method.
Per-tensor bit-width allocation via advanced sensitivity analysis and budget-constrained optimisation — no calibration data required.
Metrics
| Metric | Value |
|---|---|
| Size | 19 GB |
| Average bits | 5.1 |
| WikiText-2 PPL (median) | 8.7520 |
| PPL vs BF16 | +10.5% |
| MMLU vs BF16 | 107.3% of BF16 |
Usage
from mlx_lm import load, generate
model, tokenizer = load("baa-ai/GLM-4.7-Flash-RAM-20GB-MLX")
response = generate(model, tokenizer, prompt="Hello!", max_tokens=256)
print(response)
Quantized by baa.ai
- Downloads last month
- 545
Model size
30B params
Tensor type
BF16
·
F32 ·
U32 ·
Hardware compatibility
Log In to add your hardware
4-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for baa-ai/GLM-4.7-Flash-RAM-20GB-MLX
Base model
zai-org/GLM-4.7-Flash