readygpu

Llama 3.1 8B

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

chat

What to run it on in Germany

Fully in memory at 64k context, priced at retailers today.

Something we price holds it fully in memory, but none reaches 15 tok/s, our bar for recommending it. The table below has the speeds.

All hardware that runs it

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

HardwareQuantFitsDecodePricetok/s per 1,000
NVIDIA GeForce RTX 3090Q6_Kfits38 tok/s––
NVIDIA GeForce RTX 5090Q6_Kfits73 tok/s––
Mac Studio M4 Max 128GBQ6_Kfits23 tok/s––
NVIDIA GeForce RTX 4090Q6_Kfits41 tok/s––
NVIDIA GeForce RTX 5070 TiQ6_Kfits37 tok/s––
Mac mini M4 Pro 64GBQ6_Kfits12 tok/s––
NVIDIA GeForce RTX 5060 Ti 16GBQ6_Kfits18 tok/s––
AMD Radeon RX 7900 XTXQ6_Kfits35 tok/s––
Framework Desktop Ryzen AI Max+ 395 128GBQ6_Kfits9 tok/s––
Mac Studio M3 Ultra 512GBQ6_Kfits35 tok/s––
NVIDIA DGX Spark 128GBQ6_Kfits11 tok/s––
AMD Radeon RX 9070 XTQ5_K_Mfits25 tok/s––
GMKtec EVO-X2 Ryzen AI Max+ 395 128GBQ6_Kfits9 tok/s––
Mac Studio M3 Ultra 256GBQ6_Kfits35 tok/s––
Mac mini M4 Pro 48GBQ6_Kfits12 tok/s––
NVIDIA GeForce RTX 4060 Ti 16GBQ6_Kfits12 tok/s––
NVIDIA GeForce RTX 5080Q6_Kfits39 tok/s––
Mac Studio M4 Max 64GBQ6_Kfits23 tok/s––
NVIDIA GeForce RTX 4070 Ti SuperQ6_Kfits27 tok/s––
AMD Radeon RX 9060 XT 16GBQ5_K_Mfits12 tok/s––
Mac Studio M3 Ultra 96GBQ6_Kfits35 tok/s––
NVIDIA GeForce RTX 4080 SuperQ6_Kfits30 tok/s––
NVIDIA RTX PRO 6000 BlackwellQ6_Kfits73 tok/s––
AMD Radeon AI PRO R9700Q6_Kfits23 tok/s––
AMD Radeon RX 7800 XTQ5_K_Mfits24 tok/s––
AMD Radeon RX 7900 XTQ6_Kfits29 tok/s––
ASUS Ascent GX10 128GBQ6_Kfits11 tok/s––
Framework Desktop Ryzen AI Max+ 395 64GBQ6_Kfits9 tok/s––
Mac mini M4 32GBQ6_Kfits5 tok/s––
NVIDIA GeForce RTX 3090 TiQ6_Kfits41 tok/s––
AMD Radeon RX 7600 XTQ5_K_Mfits11 tok/s––
Intel Arc Pro B60Q6_Kfits15 tok/s––
NVIDIA Tesla P40Q6_Kfits14 tok/s––
Intel Arc A770 16GBQ5_K_Mfits20 tok/s––
NVIDIA GeForce RTX 3060 12GBQ4_K_Moffload11 tok/s––
NVIDIA GeForce RTX 5070Q4_K_Moffload14 tok/s––
NVIDIA GeForce RTX 4070 SuperQ4_K_Moffload12 tok/s––
Intel Arc B580Q4_K_Moffload10 tok/s––
NVIDIA GeForce RTX 3080 10GBQ4_K_Moffload10 tok/s––

1 other card or machine 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