My flight to Faro had no Wi-Fi for sale, so I brought a coding assistant that runs on the laptop: Qwen 3.6 35B on Ollama, driven through Claude Code on an M5 MacBook Pro.
It ran. It also answered slower than the frontier models I use every day and lost track of anything longer than a few steps. The 32K context window was the worst part: I'd cut it to that size for speed, and the model started inventing file names early. The battery fell fast enough that I watched the percentage more than the code.
This is one flight on one machine, and I didn't log numbers in the air. Your laptop, model, and task may give a different answer.
The full post has:
- a pre-flight checklist for an offline LLM
- a ten-minute battery test you run on the ground
- what a seat USB port can and can't do for a laptop
- the exact Modelfile and launch command I used
Read the full post, with the checklist and the ten-minute test: https://digitaldias.com/blog/2026-11-02-local-llm-on-a-plane/
Top comments (0)