Question 1

How much VRAM does GLM-4.6 need?

Accepted Answer

GLM-4.6 (356.8B) requires approximately 807.9205741882324 GB of VRAM at full precision (FP16), 404 GB at 8-bit quantization (Q8), and 202 GB at 4-bit quantization (Q4). The actual VRAM usage may be slightly higher due to KV cache and runtime overhead.

Question 2

What GPUs can run GLM-4.6?

Accepted Answer

We currently have no compatible GPUs indexed for GLM-4.6. This may be because the model's VRAM requirements exceed available GPUs in our database, or VRAM data hasn't been computed yet.

Question 3

What is the cheapest GPU to run GLM-4.6?

Accepted Answer

We don't currently have cloud pricing data for GPUs compatible with GLM-4.6. Cloud pricing data is updated regularly — check back soon.

Question 4

Can GLM-4.6 run on consumer hardware?

Accepted Answer

GLM-4.6 requires approximately 807.9205741882324 GB of VRAM at FP16, and even with 4-bit quantization (Q4) it needs about 202.0 GB. This exceeds the 24 GB available on the highest-end consumer GPU (RTX 4090), so GLM-4.6 requires datacenter or cloud GPUs.

Question 5

What quantization options are available for GLM-4.6?

Accepted Answer

GLM-4.6 supports the following quantization levels: FP16 (full precision): ~807.9205741882324 GB, Q8 (8-bit): ~404 GB, Q6 (6-bit): ~303 GB, Q4 (4-bit): ~202 GB. Lower quantization reduces VRAM usage at the cost of some quality loss. Q8 offers a good balance between quality and memory savings, while Q4 provides maximum compression. As a Mixture of Experts (MoE) model with 160 experts (8 active), GLM-4.6 has a larger total weight size but only activates a subset of parameters per token, which can affect effective memory usage during inference.

GLM-4.6

VRAM Requirements

Compatible GPUs

GLM-4.6 FAQ

VRAM requirements for GLM-4.6 at different quantization levels
Quantization	VRAM Required
FP16	807.9 GB
Q8	404 GB
Q6	303 GB
Q4	202 GB