Local Lab
Experiments, tool comparisons, speed notes, prompt tests, and useful failures.
Latest public items
Public CMS entries published into this workspace section.
Rapid-MLX on the Same Qwen3.6 35B A3B Model: Fast First Token, Slower Sustained Run
A Rapid-MLX benchmark using the same Qwen3.6 35B A3B model path normally loaded in LM Studio, compared with prior LM Studio and oMLX results.
The LM Studio Long-Context Test: Where a 35B Local Model Started to Hit the Wall
A practical LM Studio long-context benchmark showing how time to first token changed from 2K to 49K tokens on a 35B local model.
LM Studio vs oMLX on a MacBook Pro: The 35B Local AI Test That Actually Changed My Default
A practical benchmark of LM Studio and oMLX on a MacBook Pro running a 35B local model, including sustained generation and prefix-cache behavior.
Prompt template failure log
Prompt template failure log recorded as a reusable lab note.
MacBook Air local model baseline
MacBook Air local model baseline recorded as a reusable lab note.
Small model speed baseline
A repeatable lab format for testing whether small local models are fast enough.