The surprising part is not the 27B model - it is the hardware claim
Qwen 3.8 27B reportedly running locally on an 8GB RTX 4060 via Unsloth IQ4_XS, with 64k context, 14.6GB on disk, ~150 tok/s prefill, and ~5 tok/s decode with native MTP.
Qwen 3.8-27B reads your coaching calls and writes the notes for you.
Alibaba dropped it on August 14th, and Token Harbor just added a free route to run it. No GPU. No 55GB download. You paste a prompt and go. → 1 million tokens of context, so feed it your whole document
Unlock Qwen 3.8's Secrets_ Enable Plugin & Test AI Knowledge!
Qwen 3.8 Flash can read a whole month of your notes in one go.
Alibaba dropped it on August 26th, and the numbers are strange. 125 billion parameters. But only 6 billion turn on per token. Big brain. Tiny bill. → 1 million token context window → Text, images, and video all
You have reached the end of the archive
All of qwen38