token optimization

2 articles tagged with this topic

Qwen Users Find Token Hack: Merging Two 27B Models Cuts Cost, Holds Performance

Qwen community users found merging two 27B model weights cuts token output while preserving performance—a signal model merging is becoming a budget mo

Sep 282 min read

MiniMaxOpenRouter

LLMs Now Think in 'Caveman Language' — Not Getting Dumber, Just Saving Money

Space Bunny Alpha compresses chain-of-thought into 'caveman language'—by design, not bug. Part of an industry push to cut token costs.

Sep 242 min read