Question 1

How much VRAM does CodeLlama-70b-Instruct-hf need?

Accepted Answer

CodeLlama-70b-Instruct-hf (70B) requires approximately 154.17540893554687 GB of VRAM at full precision (FP16), 77.1 GB at 8-bit quantization (Q8), and 38.5 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 CodeLlama-70b-Instruct-hf?

Accepted Answer

We have 5 compatible GPUs indexed for CodeLlama-70b-Instruct-hf, including MI300X, B200, B100, 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 CodeLlama-70b-Instruct-hf?

Accepted Answer

The cheapest GPU to run CodeLlama-70b-Instruct-hf in the cloud is the MI300X at $1.99/hr. Other affordable options include B200 ($5.49/hr). Prices reflect the lowest on-demand rate across cloud providers.

Question 4

Can CodeLlama-70b-Instruct-hf run on consumer hardware?

Accepted Answer

CodeLlama-70b-Instruct-hf requires approximately 154.17540893554687 GB of VRAM at FP16, and even with 4-bit quantization (Q4) it needs about 38.5 GB. This exceeds the 24 GB available on the highest-end consumer GPU (RTX 4090), so CodeLlama-70b-Instruct-hf requires datacenter or cloud GPUs.

Question 5

What quantization options are available for CodeLlama-70b-Instruct-hf?

Accepted Answer

CodeLlama-70b-Instruct-hf supports the following quantization levels: FP16 (full precision): ~154.17540893554687 GB, Q8 (8-bit): ~77.1 GB, Q6 (6-bit): ~57.8 GB, Q4 (4-bit): ~38.5 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.

CodeLlama-70b-Instruct-hf

VRAM Requirements

Compatible GPUs

CodeLlama-70b-Instruct-hf FAQ

GPU	VRAM	Best Quant	From	Rent
MI300X	192 GB	FP16	$1.99/hr	Rent
B200	192 GB	FP16	$5.49/hr	Rent
B100	192 GB	FP16	—
MI325X	256 GB	FP16	—
MI350X	288 GB	FP16	—

VRAM requirements for CodeLlama-70b-Instruct-hf at different quantization levels
Quantization	VRAM Required
FP16	154.2 GB
Q8	77.1 GB
Q6	57.8 GB
Q4	38.5 GB