Can You Run Meta-Llama-3.1-8B? Every GPU, Tested

A model that doesn't fit doesn't run slowly, it doesn't run. Here's every card from our bench, sorted by measured speed.

Weights: bartowski/Meta-Llama-3.1-8B-Instruct-GGUF

GPUs that run Meta-Llama-3.1-8B (11)

GPUSpeedVRAMBasis
NVIDIA B300293 tok/s288GB✓ Measured
NVIDIA B200274.17 tok/s192GB✓ Measured
NVIDIA H200262.93 tok/s141GB✓ Measured
NVIDIA H100 80GB HBM3261.7 tok/s80GB✓ Measured
NVIDIA RTX PRO 6000 Blackwell Workstation Edition237.69 tok/s96GB✓ Measured
NVIDIA A100 80GB SXM4164.24 tok/s80GB✓ Measured
NVIDIA A100 40GB SXM4156.68 tok/s40GBEst.
NVIDIA L40S135.46 tok/s48GB✓ Measured
NVIDIA A10G86.69 tok/s24GB✓ Measured
NVIDIA L450.36 tok/s24GB✓ Measured
NVIDIA T437.8 tok/s16GB✓ Measured

Quick answers

What is the fastest GPU for Meta-Llama-3.1-8B?

NVIDIA B300 is the fastest card we've measured running Meta-Llama-3.1-8B, at 293 tok/s.

What is the cheapest GPU that can run Meta-Llama-3.1-8B?

NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs Meta-Llama-3.1-8B, at 37.8 tok/s.

← All models · AI GPU rankings · How we benchmark