作者： Jakub Rusinowski · 最后更新： 2024年7月23日
Model library → Llama 3.1 Family → Llama 3.1 8B Instruct
The go-to model for consumer hardware. Excellent reasoning, tool use, and multilingual support.
Llama 3.1 8B Instruct needs about 6 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
| Parameters | 8 Billion |
| Context window | 128,000 |
| Architecture | Dense Decoder-only |
| Provider | Meta |
| Licence | Llama Community License |
| Specified at | Q4_K_M |
| System RAM | 16 GB |
| Record updated | 2024-07-23 |
Llama Community — commercial use permitted. Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
Modelled on a reference NVIDIA RTX 4090 (24 GB), at 8K context. Speed figures are ESTIMATES from the memory-bandwidth roofline described on the methodology page, not benchmarks we ran — rows marked measured come from published or reader-submitted runs. VRAM here includes the KV cache, so it reads higher than the headline figure above, which does not.
| Quant | Weights | VRAM needed | Est. speed | Fit on 24 GB |
|---|---|---|---|---|
| Q2_K | 2.6 GB | 4.5 GB | ~155 tok/s (est.) | Fits comfortably |
| Q3_K_M | 3.4 GB | 5.3 GB | ~134 tok/s (est.) | Fits comfortably |
| Q4_K_M | 4.8 GB | 6.7 GB | ~107 tok/s (est.) | Fits comfortably |
| Q5_K_M | 5.7 GB | 7.5 GB | ~96 tok/s (est.) | Fits comfortably |
| Q6_K | 6.6 GB | 8.4 GB | ~86 tok/s (est.) | Fits comfortably |
| Q8_0 | 8.5 GB | 10.4 GB | ~70 tok/s (est.) | Fits comfortably |
| F16 | 16.0 GB | 17.9 GB | ~41 tok/s (est.) | Fits comfortably |
Want the memory numbers alone, at every quantization level and your own context length? Use the Llama 3.1 8B Instruct VRAM calculator.
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
The cheapest catalogued GPU that runs Llama 3.1 8B Instruct is the Intel Arc B570 (10 GB).
Install Ollama, then run:
ollama run llama3.1Weights on Hugging Face: meta-llama/Meta-Llama-3.1-8B-Instruct.
Pick a quantization and open it in LM Studio, Ollama, or Jan, or download the raw .gguf file directly. Quant list and sizes resolved from Hugging Face.
| Quant | Size | Download (.gguf) |
|---|---|---|
| Q3_K_M | 3.41 GB (est.) | Meta-Llama-3.1-8B-Instruct-Q3_K_M.gguf |
| Q4_K_M | 4.83 GB (est.) | Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf |
| Q5_K_M | 5.67 GB (est.) | Meta-Llama-3.1-8B-Instruct-Q5_K_M.gguf |
| Q6_K | 6.56 GB (est.) | Meta-Llama-3.1-8B-Instruct-Q6_K.gguf |
| Q8_0 | 8.50 GB (est.) | Meta-Llama-3.1-8B-Instruct-Q8_0.gguf |
Download in LM Studio: lms get bartowski/Meta-Llama-3.1-8B-Instruct-GGUF
Want this model on your phone? You can run it on your desktop with LM Studio and chat from your iPhone or iPad over an encrypted link — see Run LM Studio Models on Your Phone (LM Link).
Best for: chat, rag, agents.
← All Llama 3.1 Family models | VRAM calculator | Check your own hardware