Live site · by chaoliangUNSW

Jev-Style 0.8B Decision v3

Jev-style decisions on a laptop, 0.53 GB in 4-bit.

chaoliangUNSW's GGUF builds of a 0.8B decision model for llama.cpp. The 4-bit build gives the same decision as full precision on 240 of 240 parity rows. The companion jev-style package serves it behind a systemone-compatible API and adds six agent skills. On tasks outside its training data it beats the best official Laya checkpoint on Banking77 and 37 held-out MASSIVE languages.