Announcement
TwIL-LM3-Pro is claimed to score 95.4% on BIG-Bench Hard’s logic subset
A user says the open-source model has 3.66 billion parameters and is small enough to run locally on a laptop.
TLDR
A user claims TwIL-LM3-Pro scores 95.4% on BIG-Bench Hard’s logic subset with 3.66 billion parameters. They say it can run locally on a laptop and describe specialized post-training as a way to make small models for specific tasks.
Combined views
92.3K
5 Sources, first seen 7h ago
230 likes
