@dgm.america: How to pick a local AI model that actually fits your Mac, in four steps. Save this one. ① Measure the desk: Apple menu → About This Mac → Memory. Budget about ¾ (32 GB → ~24 GB). ② Read the name: in Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf, two parts decide the fit. 30B = size (all of it sits in memory). Q4_K_M = quantization, about 4 bits per number instead of 16. ③ Do the math: at Q4, about 0.6 GB per billion parameters. 8B ≈ 5 GB, 30B ≈ 18.6 GB, 70B ≈ 42 GB. ④ Leave room to work: the chat itself (the KV cache) needs memory too. For Qwen3 30B-A3B, about 0.1 GB per 1,000 tokens. On a 32 GB Mac: Q8 ✗, Q5 ✗, Q4_K_M tight, Q3 fits. osFoundry does this check for you and badges each model: fits, tight, or no. #LocalLLM #LocalAI #Ollama #LMStudio #MacAI

966

There are no more comments for this video.

To see more videos from user @dgm.america, please go to the Tikwm homepage.