AIHacker Newsabout 2 hours ago
Jeff: 0.8B Decision Models Trained at Home
6 min read
An independent project released Jeff, fine-tunes of Qwen3.5 and Gemma 4 that return calibrated probabilities for zero-shot classification in about 22 ms on an RTX PRO 6000 and 28 ms on an Apple M4 Max. Models are on Hugging Face under Apache 2.0, with training done entirely on local hardware.
Level
Hype check
Tap to vote and see what everyone thinks.
