Tried DeepSeek R1 distilled Qwen 14B Q8 1m context window MLX on LM Studio on 14" Macbook Pro M4 Pro 14c/20gpu, 48GB, 1TB store it's taking 3+ hrs to just process the UK Online Safety Act pdf files only 40% processed 😆 Google Gemini takes ~25 seconds & reports PDF ~500K tokens