So you bought the 128GB MacBook Pro. Now the question is not, “Which local model gets the highest TPS?” It is: which setup can I actually trust to get the job done? This is the local coding stack I’d start with: Qwen 3.6, dense 27B, Q6 quant, MLX server, 8192 output tokens, 20GB prompt cache, and deterministic decoding. If Anthropic’s success story tells us anything, it is that once you figure out coding, you can expand into almost anything else. Local models stop being a hobby when they can finish the patch.

oMLX hits 47 tokens per second on a base M2 MacBook Pro by offloading context to the SSD. We explore how native MLX features achieve 3x faster generation than LM Studio in our latest test.

This is where we are right now. And i’m not gonna lie it feels pretty magical 🧚‍♀️ Qwen3.6 27B running inside of Pi coding agent via Llama.cpp on the MacBook Pro For non-trivial tasks on the @huggingface codebases, this feels very, very close to hitting the latest Opus in Claude Code, or whatever shiny monopolistic closed source API of the day is. In full airplane mode. Most people haven’t realized this yet. If you have, it means you have a huge headstart to what I call the second revolution of AI. Powerful local models for efficiency, security, privacy, sovereignty 🔥

1d 13h 20m, 3,596,831 tokens. Goal achieved? Not quite. It was a hard problem. The agent tried its best and went through 20 full model/eval rounds. In the end, the agent talked itself out of the original contract and declared the goal achieved. I probably would have stopped it anyway, since I could also see from the sidecar that it was struggling. Still, it was a good experiment. My 14" MacBook Pro held up well under a sustained run, with no throttling or heating issue. Qwen3.6 35B A3B OptiQ 4-bit running locally on MLX also held up well. It generated thousands of training data samples, averaging around 50 tps with reasonably good quality. Very impressive. DeepSeek 4 Pro was a good teacher for the training, though there are still areas for improvement. The end result: we LoRAed an expert model, Qwen3-4B-Instruct-2507 + MLX LoRA. We produced a compact 56 MB LoRA adapter on a 4B Qwen base that reaches ~59% three-way decision agreement on the original eval slice, ~91% violation recall, and ~98% valid JSON, but with a high false-positive rate. It is deployable, but probably not quite usable yet. Still, it gives me a clear direction for where to go next. I’ll write more about the whole process later. Stay tuned.