inferencerlabs/GLM-5-MLX-5.6bit-INF

NOTICE

This model has been superseded by the higher quality GLM5.1-Q4.8-INF version available here

See GLM-5 MLX-INF in action - demonstration video

Tested across a M3 Ultra 512GB and M4 Max 128GB using Inferencer distributed compute

• Distributed inference ~12.5 tokens/s @ 1000 tokens
• Memory usage: ~444 GiB / 49 GiB

Q5.6-INF typically achieves near lossless accuracy to higher quants in our coding test, using an early version of the data-agnostic INF method tuned to fit within a 640GB memory budget

Quantization	Perplexity	Token Accuracy	Missed Divergence
Q4.5	1.32812	90.50%	26.44%
Q5.5	1.23437	95.40%	16.03%
Q5.6-INF	1.21875	96.75%	12.35%
Q6.5	1.21875	96.85%	12.55%
Q8.5	1.21875	97.65%	9.92%
Q9	1.21093	97.95%	9.61%
Base	1.20312	100.0%	0.000%

Quantized with a modified version of MLX

For more details see demonstration video or visit GLM-5.

Disclaimer

We are not the creator, originator, or owner of any model listed. Each model is created and provided by third parties. Models may not always be accurate or contextually appropriate. You are responsible for verifying the information before making important decisions. We are not liable for any damages, losses, or issues arising from its use, including data loss or inaccuracies in AI-generated content.