4 ms·
What model are you running with 64GB of VRAM (equivalent)? I doubt most users are doing that. Looking at their documentation, the default path for openclaw seem
by rushcar 7mo ago
What model are you running with 64GB of VRAM (equivalent)? I doubt most users are doing that. Looking at their documentation, the default path for openclaw seems to be a 3P API for the model.
- lwhi 7mo agoIt doesn't matter what 'most users' are doing. On a 64 GB Apple silicon Mac mini you can natively host mid sized and some larger quantised local models .. using Ollama. For example: Qwen3-Coder (32B), GLM-4.7 (or GLM-4 Variants), Devstral-24B / Mistral Large (Quantized)