Spaces:

---

webAI-Official

/

TwIL-LM3

like 1

Running on Zero

The Space README recommends 2048 or more max new tokens to reduce reasoning
truncation, while the app defaults to 1024 and explains in source that
anonymous visitors have a smaller quota.

Would you welcome a README-only clarification that states the tradeoff
directly? A minimal wording would be:

“The Space defaults to 1024 max new tokens to accommodate the smaller quota
for anonymous visitors; use 2048 or higher when quota permits to reduce the
chance of truncating the reasoning block.”

This would not change the application default or model behavior. I have not
opened a pull request or run the model.

Prepared with OpenAI Codex assistance and reviewed by me.