🔌
New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
// LOCAL-GENAI HARDWARE
Hardware for local LLMs
GPUs and unified-memory systems ranked for running downloadable models — memory, bandwidth, max model, estimated tokens/sec.
Missing something? Add it.
10 devices
Consumer AMD Radeon RX 7900 XTX
AMD
24GB
Memory
960 GB/s
Bandwidth
32.5B
Max model Q4
135-165
8B Q4 tok/s
Consumer AMD Radeon RX 7900 XT
AMD
20GB
Memory
800 GB/s
Bandwidth
26B
Max model Q4
112.5-137.5
8B Q4 tok/s
Consumer AMD Radeon RX 9070 XT
AMD
16GB
Memory
644 GB/s
Bandwidth
19.5B
Max model Q4
90.6-110.7
8B Q4 tok/s
Consumer AMD Radeon RX 7800 XT
AMD
16GB
Memory
624 GB/s
Bandwidth
19.5B
Max model Q4
87.8-107.3
8B Q4 tok/s
Consumer AMD Radeon RX 6900 XT
AMD
16GB
Memory
512 GB/s
Bandwidth
19.5B
Max model Q4
72-88
8B Q4 tok/s
Consumer AMD Radeon RX 6800 XT
AMD
16GB
Memory
512 GB/s
Bandwidth
19.5B
Max model Q4
72-88
8B Q4 tok/s
Prosumer AMD Radeon Pro W7900 48GB
AMD
48GB
Memory
864 GB/s
Bandwidth
71B
Max model Q4
121.5-148.5
8B Q4 tok/s
AM Prosumer
AMD Ryzen AI Max+ 395
AMD
128GB
Memory
256 GB/s
Bandwidth
195B
Max model Q4
36-44
8B Q4 tok/s
AM Enterprise
AMD Instinct MI325X
AMD
256GB
Memory
6000 GB/s
Bandwidth
410B
Max model Q4
843.8-1031.3
8B Q4 tok/s
AM Enterprise
AMD Instinct MI300X 192GB
AMD
192GB
Memory
5300 GB/s
Bandwidth
305B
Max model Q4
745.3-910.9
8B Q4 tok/s
How to read this
Two numbers dominate local inference: memory (how big a model + context fits) and bandwidth (token speed). Click any hardware for the full size×quant matrix, max usable context and measured reports, or browse runtimes.