
Run LLM Locally – which open model fits your GPU or Mac, and how fast
New to this?Read our local LLM guide for beginnersFind the LLM that fits your machine
以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。
New to this?Read our local LLM guide for beginnersFind the LLM that fits your machine
Enter your GPU or Mac to see which models run, with the memory they need and an estimated speed. Or pick a model and see what hardware runs it.
What hardware do I have?See which models run on your PC or Mac.What model do I want to run?Find the hardware a model needs.Build me a compatible machineGet PC and Mac configurations for a target.
Machine typePC with a GPUMac (Apple Silicon)Unified-memory PCPC without a GPU
Graphics cardAMD Radeon PRO W7900 · 48 GBAMD RX 7900 XTX · 24 GBAMD RX 7900 XT · 20 GBAMD RX 9070 XT · 16 GBAMD RX 9070 · 16 GBAMD RX 7800 XT · 16 GBAMD RX 7900 GRE · 16 GBAMD RX 6800 XT · 16 GBAMD RX 7600 XT · 16 GBIntel Arc A770 16 GBIntel Arc B580 · 12 GBNVIDIA RTX PRO 6000 Blackwell · 96 GBNVIDIA H100 PCIe · 80 GBNVIDIA A100 80GBNVIDIA RTX 6000 Ada · 48 GBNVIDIA RTX A6000 · 48 GBNVIDIA RTX 5090 · 32 GBNVIDIA RTX 4090 · 24 GBNVIDIA RTX 3090 Ti · 24 GBNVIDIA RTX 3090 · 24 GBNVIDIA Tesla P40 · 24 GBNVIDIA RTX 5080 · 16 GBNVIDIA RTX 5070 Ti · 16 GBNVIDIA RTX 4080 SUPER · 16 GBNVIDIA RTX 4080 · 16 GBNVIDIA RTX 4070 Ti SUPER · 16 GBNVIDIA RTX 5060 Ti 16 GBNVIDIA RTX 4060 Ti 16 GBNVIDIA RTX 3080 Ti · 12 GBNVIDIA RTX 5070 · 12 GBNVIDIA RTX 4070 Ti · 12 GBNVIDIA RTX 4070 SUPER · 12 GBNVIDIA RTX 4070 · 12 GBNVIDIA RTX 3060 12 GBNVIDIA RTX 2080 Ti · 11 GBNVIDIA RTX 3080 · 10 GBNVIDIA RTX 5060 Ti 8 GBNVIDIA RTX 5060 · 8 GBNVIDIA RTX 3070 · 8 GBNVIDIA RTX 4060 Ti 8 GBNVIDIA RTX 4060 · 8 GB
System RAM (GB)
Advanced options8K tokens · KV cache f16The defaults suit most uses. You can skip this section.
How much text the model keeps in memory. 8K is enough for chat; long documents need 32K or more. What is this?
Leave f16 if unsure. q8_0 roughly halves the conversation's memory, q4_0 nearly quarters it; the option must also be enabled in your app. What is this?
ProcessorI don't know / otherIntel Core i5-13600KIntel Core i7-14700KIntel Core i9-14900KIntel Core Ultra 7 265KIntel Core Ultra 9 285KAMD Ryzen 5 5600XAMD Ryzen 5 7600AMD Ryzen 7 5800X3DAMD Ryzen 7 7800X3DAMD Ryzen 7 9700XAMD Ryzen 7 9800X3DAMD Ryzen 9 7950XAMD Ryzen 9 9950XAMD Ryzen Threadripper 7970X
Used to estimate RAM speed, which only matters for the part of the model that does not fit on the graphics card.
Memory = weights + KV cache + runtime overhead + safety margin, from each model's architecture. Speeds come from real benchmarks when they exist; otherwise they are estimated from memory bandwidth and labelled as estimates, with a confidence.
正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。