🔌
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.
6 devices
Consumer Intel Arc A770 16GB
Intel
16GB
Memory
560 GB/s
Bandwidth
19.5B
Max model Q4
72.5-90
8B Q4 tok/s
Consumer Intel Arc B580 12GB
Intel
12GB
Memory
456 GB/s
Bandwidth
14.5B
Max model Q4
59-73.3
8B Q4 tok/s
Prosumer Intel Arc Pro B70
Intel
32GB
Memory
608 GB/s
Bandwidth
45B
Max model Q4
78.7-97.7
8B Q4 tok/s
Prosumer Intel Arc Pro B60
Intel
24GB
Memory
456 GB/s
Bandwidth
32.5B
Max model Q4
59-73.3
8B Q4 tok/s
Prosumer Intel Arc Pro B50
Intel
16GB
Memory
224 GB/s
Bandwidth
19.5B
Max model Q4
29-36
8B Q4 tok/s
SME Intel Gaudi 3
Intel
128GB
Memory
3700 GB/s
Bandwidth
199.5B
Max model Q4
479-594.6
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.