Llama 3.3 70B

70.6B parameters
reasoningcoding
Parameters
70.6B
Context Length
128K
Architecture
llama
Release Date
2024

Available Quantizations

QuantizationFile SizeQuality
Q2_K26.5 GBLower
Q3_K_M33.5 GBLower
Q4_K_M41.0 GBGood
Q5_K_M47.5 GBGood
Q8_075.0 GBHigh

Can it run? (Q4_K_M quantization)

GeForce RTX 4090
Won't run
~0 tok/s • 24GB VRAM
GeForce RTX 4070
Won't run
~0 tok/s • 12GB VRAM
GeForce RTX 3060 12GB
Won't run
~0 tok/s • 12GB VRAM
Radeon RX 7900 XTX
Won't run
~0 tok/s • 24GB VRAM
Check with your specific GPU →