PracticlySpeaking

• Offload settings for unsloth/Gemma-4 on Apple Silicon?

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 2 comments

• Ollama 0.19 with MLX is the real deal

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 3 comments

• New - Apple Neural Engine (ANE) backend for llama.cpp

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 24 comments

• GPT-oss-120b - What is up with GPU Offload setting (LM Studio / Mac)

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 4 comments

• MacOS 26.2 to add full 'Neural Accelerator' support for M5 chips

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 32 comments

• Is MLX working with new M5 matmul yet?

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 20 comments

• Anyone with a 64GB Mac and unsloth gpt-oss-120b — Will it load with full GPU offload?

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 15 comments

• Getting models to load in LM Studio [64GB Mac]

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 11 comments

• Is it just me, or is LM Studio really pushing the new gpt-oss?

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 55 comments

• Qwen3-Coder-30B-A3B MLX-DWQ variants like lr9e8

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 5 comments

• What would you run with 128GB RAM instead of 64GB? (Mac)

  Posted by PracticlySpeaking@reddit | LocalLLaMA | View on Reddit | 39 comments