Question 1

How much VRAM does Jamba-v0.1 need?

Accepted Answer

Jamba-v0.1 (51.6B) requires approximately 205.80528259277344 GB of VRAM at full precision (FP16), 102.9 GB at 8-bit quantization (Q8), and 51.5 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 Jamba-v0.1?

Accepted Answer

We have 2 compatible GPUs indexed for Jamba-v0.1, including MI325X, MI350X. Compatibility is determined by whether the GPU's VRAM can fit the model's memory footprint, potentially using quantization to reduce memory usage.

Question 3

What is the cheapest GPU to run Jamba-v0.1?

Accepted Answer

We don't currently have cloud pricing data for GPUs compatible with Jamba-v0.1. Cloud pricing data is updated regularly — check back soon.

Question 4

Can Jamba-v0.1 run on consumer hardware?

Accepted Answer

Jamba-v0.1 requires approximately 205.80528259277344 GB of VRAM at FP16, and even with 4-bit quantization (Q4) it needs about 51.5 GB. This exceeds the 24 GB available on the highest-end consumer GPU (RTX 4090), so Jamba-v0.1 requires datacenter or cloud GPUs.

Question 5

What quantization options are available for Jamba-v0.1?

Accepted Answer

Jamba-v0.1 supports the following quantization levels: FP16 (full precision): ~205.80528259277344 GB, Q8 (8-bit): ~102.9 GB, Q6 (6-bit): ~77.2 GB, Q4 (4-bit): ~51.5 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 16 experts (2 active), Jamba-v0.1 has a larger total weight size but only activates a subset of parameters per token, which can affect effective memory usage during inference.

Jamba-v0.1

VRAM Requirements

Compatible GPUs

Jamba-v0.1 FAQ

VRAM requirements for Jamba-v0.1 at different quantization levels
Quantization	VRAM Required
FP16	205.8 GB
Q8	102.9 GB
Q6	77.2 GB
Q4	51.5 GB

Top 10 compatible GPUs for Jamba-v0.1, sorted by cheapest price
GPU	VRAM	Best Quant	From	Rent
MI325X	256 GB	FP16	—
MI350X	288 GB	FP16	—