readygpu

Llama 3.1 8B

8B parameters · 128k context · llama3.1 · released 2024-07-18

chat

What to run it on in the US

Fully in memory at 128k context, priced at retailers today.

All hardware that runs it

Everything we price, best value first. Two cards split the model by layers.

HardwareQuantFitsDecodePricetok/s per 1,000
Intel Arc Pro B60Q5_K_Mfits10 tok/s$65015.4
AMD Radeon RX 7900 XTXQ5_K_Mfits23 tok/s$1,53315.0
NVIDIA GeForce RTX 3090Q5_K_Mfits25 tok/s$1,66015.2
AMD Radeon AI PRO R9700Q6_Kfits15 tok/s$1,7008.7
NVIDIA GeForce RTX 3090 TiQ5_K_Mfits27 tok/s$2,06013.3
NVIDIA GeForce RTX 4090Q5_K_Mfits27 tok/s$3,2808.3
NVIDIA GeForce RTX 5090Q6_Kfits47 tok/s$7,2006.5
Mac Studio M4 Max 128GBQ6_Kfits15 tok/s––
Mac mini M4 Pro 64GBQ6_Kfits8 tok/s––
Framework Desktop Ryzen AI Max+ 395 128GBQ6_Kfits6 tok/s––
Mac Studio M3 Ultra 512GBQ6_Kfits22 tok/s––
NVIDIA DGX Spark 128GBQ6_Kfits7 tok/s––
GMKtec EVO-X2 Ryzen AI Max+ 395 128GBQ6_Kfits6 tok/s––
Mac Studio M3 Ultra 256GBQ6_Kfits22 tok/s––
Mac mini M4 Pro 48GBQ6_Kfits8 tok/s––
Mac Studio M4 Max 64GBQ6_Kfits15 tok/s––
Mac Studio M3 Ultra 96GBQ6_Kfits22 tok/s––
NVIDIA RTX PRO 6000 BlackwellQ6_Kfits47 tok/s––
ASUS Ascent GX10 128GBQ6_Kfits7 tok/s––
Framework Desktop Ryzen AI Max+ 395 64GBQ6_Kfits6 tok/s––
NVIDIA Tesla P40Q5_K_Mfits9 tok/s––
NVIDIA GeForce RTX 3060 12GBQ4_K_Moffload4 tok/s$36010.6
Intel Arc B580Q4_K_Moffload4 tok/s$37010.3
Intel Arc A770 16GBQ4_K_Moffload5 tok/s$40013.0
NVIDIA GeForce RTX 3080 10GBQ4_K_Moffload4 tok/s$4697.9
AMD Radeon RX 9060 XT 16GBQ4_K_Moffload5 tok/s$5438.3
AMD Radeon RX 7800 XTQ4_K_Moffload6 tok/s$5999.2
AMD Radeon RX 7600 XTQ4_K_Moffload4 tok/s$6956.2
NVIDIA GeForce RTX 4070 SuperQ4_K_Moffload4 tok/s$6995.9
NVIDIA GeForce RTX 4060 Ti 16GBQ4_K_Moffload5 tok/s$7706.0
NVIDIA GeForce RTX 5060 Ti 16GBQ4_K_Moffload5 tok/s$7727.0
NVIDIA GeForce RTX 5070Q4_K_Moffload4 tok/s$7765.4
AMD Radeon RX 7900 XTQ4_K_Moffload10 tok/s$80012.3
AMD Radeon RX 9070 XTQ4_K_Moffload6 tok/s$8156.9
NVIDIA GeForce RTX 4070 Ti SuperQ4_K_Moffload6 tok/s$1,0135.8
NVIDIA GeForce RTX 5070 TiQ4_K_Moffload6 tok/s$1,1445.5
NVIDIA GeForce RTX 4080 SuperQ4_K_Moffload6 tok/s$1,1995.0
NVIDIA GeForce RTX 5080Q4_K_Moffload6 tok/s$1,7383.6

2 other cards and machines we price cannot hold it at this context.

Memory needed

Weights, KV cache and runtime buffers at each context length (KV cache at fp16).

QuantFile@8k@32k@64k@128k
Q4_K_Mgguf4.6 GB6.3 GB9.5 GB13.7 GB22.0 GB
mlx-4bitmlx4.2 GB6.0 GB9.1 GB13.3 GB21.7 GB
Q5_K_Mgguf5.3 GB7.1 GB10.2 GB14.4 GB22.8 GB
Q6_Kgguf6.1 GB7.9 GB11.0 GB15.2 GB23.6 GB

Run it

Copy a command for the runtime you use.

  • Q4_K_M
    • Ollamaollama run hf.co/unsloth/Llama-3.1-8B-Instruct-GGUF:Q4_K_M
    • LM Studiolms get unsloth/Llama-3.1-8B-Instruct-GGUF@q4_k_m
    • llama.cppllama-server -hf unsloth/Llama-3.1-8B-Instruct-GGUF:Q4_K_M
  • mlx-4bit
    • LM Studiolms get mlx-community/Llama-3.1-8B-Instruct-4bit
    • MLXmlx_lm.server --model mlx-community/Llama-3.1-8B-Instruct-4bit
  • Q5_K_M
    • Ollamaollama run hf.co/unsloth/Llama-3.1-8B-Instruct-GGUF:Q5_K_M
    • LM Studiolms get unsloth/Llama-3.1-8B-Instruct-GGUF@q5_k_m
    • llama.cppllama-server -hf unsloth/Llama-3.1-8B-Instruct-GGUF:Q5_K_M
  • Q6_K
    • Ollamaollama run hf.co/unsloth/Llama-3.1-8B-Instruct-GGUF:Q6_K
    • LM Studiolms get unsloth/Llama-3.1-8B-Instruct-GGUF@q6_k
    • llama.cppllama-server -hf unsloth/Llama-3.1-8B-Instruct-GGUF:Q6_K