Sam Witteveen · 2026-10-04 · notable
Sam Witteveen — 'Which is The Best Qwen3.8-27B?'
Sam Witteveen's 4 October 2026 video compares three Qwen3.8-27B fine-tunes that cut thinking tokens — ThinkingCap, Swift 1.5 and QwenPi — and tests them on coding, logic, math and SVG tasks.
Sam Witteveen tests which Qwen3.8-27B fine-tune thinks less without losing accuracy.
What is it?
'Which is The Best Qwen3.8-27B?' went up on Sam Witteveen's channel on 4 October 2026. The video compares three community fine-tunes of Qwen3.8-27B — BottleCap AI's ThinkingCap, UkisAI's Swift 1.5 and QwenPi — that all aim to cut reasoning tokens while keeping accuracy.
How does it work?
Chapters explain why reasoning models think too long and how Qwen3.8's reasoning-effort setting works, then take each fine-tune in turn. A demo and a round of coding, logic, math and SVG tests close the video. The makers' own claims differ: ThinkingCap reports 37.2% fewer thinking tokens, Swift 1.5 reports 58.5% fewer on some tasks, and QwenPi is trained on Pi coding-agent sessions.
Why does it matter?
Fewer thinking tokens means faster and cheaper answers from the same 27B model on local hardware. A side-by-side test helps builders pick a fine-tune for their own use case instead of trusting each maker's benchmarks.
Who is it for?
developers running Qwen3.8-27B locally