Llama 3.1 8B Instruct Legacy
Files and memory at 8K context
No verified quantized file for the requested formats.
Sizes are exact file bytes (text weights; projectors listed separately). “+” = recurrent state or other terms unknown. Smallest fit adds 0.5 GiB spare; Macs use 75% of memory. Legacy model: sizes on device pages are estimated from parameter count.
Quality
Artificial Analysis Intelligence Index
Unknown
Arena text rating
Unknown
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 33
| Hardware | Quant | Backend | Prompt t/s | Gen t/s | Date | Source |
|---|---|---|---|---|---|---|
| NVIDIA DGX Spark | Q4_K_M | Ollama | — | 38 | 2025-10-23 | ollama.com |
| NVIDIA DGX Spark | Q8_0 | Ollama | — | 25.2 | 2025-10-23 | ollama.com |
| NVIDIA DGX Spark | Q4_K_M | Ollama | — | 43.2 | 2025-10-13 | docs.google.com |
| NVIDIA DGX Spark | Q8_0 | Ollama | — | 28.5 | 2025-10-13 | docs.google.com |
| RTX Pro 6000 Blackwell Edition | Q4_K_M | Ollama | — | 202 | 2025-10-13 | docs.google.com |
| RTX Pro 6000 Blackwell Edition | Q8_0 | Ollama | — | 144 | 2025-10-13 | docs.google.com |
| GeForce RTX 5090 | Q4_K_M | Ollama | — | 200 | 2025-10-13 | docs.google.com |
| GeForce RTX 5090 | Q8_0 | Ollama | — | 145 | 2025-10-13 | docs.google.com |
| GeForce RTX 5080 | Q4_K_M | Ollama | — | 135 | 2025-10-13 | docs.google.com |
| GeForce RTX 5080 | Q8_0 | Ollama | — | 90.7 | 2025-10-13 | docs.google.com |
| NVIDIA GeForce RTX 3060 Laptop GPU | Q4_K_M | llama.cpp Vulkan | 1,061 | 9.1 | 2025-11-18 | llama.cpp #10879 |
| AMD Radeon(TM) Graphics | Q4_K_M | llama.cpp Vulkan | 230 | 12.1 | 2025-11-18 | llama.cpp #10879 |
| NVIDIA DGX Spark | NVFP4 | TensorRT-LLM | 10,257 | 38.6 | 2025-10-24 | developer.nvidia.com |
| NVIDIA DGX Spark | FP8 | SGLang | 7,991 | 20.5 | 2025-10-13 | docs.google.com |
| RTX Pro 6000 Blackwell Edition | FP8 | SGLang | 38,744 | 144 | 2025-10-13 | docs.google.com |
| GeForce RTX 5090 | FP8 | SGLang | 21,956 | 138 | 2025-10-13 | docs.google.com |
| Mac Studio M1 Max | Q4_K_M | Ollama | — | 42.3 | 2025-10-13 | docs.google.com |
| Mac Studio M1 Max | Q8_0 | Ollama | — | 33.2 | 2025-10-13 | docs.google.com |
| Mac Mini M4 Pro | Q4_K_M | Ollama | — | 34 | 2025-10-13 | docs.google.com |
| Mac Mini M4 Pro | Q8_0 | Ollama | — | 26.1 | 2025-10-13 | docs.google.com |
| NVIDIA GeForce RTX 3090 | Q4_0 | llama.cpp CUDA | 5,038 | 145 | 2025-09-14 | llama.cpp #15013 |
| NVIDIA GeForce RTX 3090 | Q4_0 | llama.cpp CUDA | 4,359 | — | 2025-09-14 | llama.cpp #15013 |
| NVIDIA GeForce RTX 3090 | Q4_0 | llama.cpp CUDA | 3,836 | — | 2025-09-14 | llama.cpp #15013 |
| NVIDIA GeForce RTX 3060 | Q4_0 | llama.cpp CUDA | 2,055 | 66.6 | 2025-09-14 | llama.cpp #15013 |
| NVIDIA GeForce RTX 3060 | Q4_0 | llama.cpp CUDA | 1,800 | — | 2025-09-14 | llama.cpp #15013 |
Reported by the linked sources, not reproduced by LocalFit. Showing 25 of 33. All rows in the speed database