Underdog Saluki 27B Launch: Comparing 2-Bit GGUFs and Testing Tool Calling in llama.cpp
This guide examines building Saluki based on Qwen3.8-27B and comparing it with the GSQ-RCO version of ISTA—from GGUF structure and memory consumption to behavior on BFCL tasks when run with standard llama.cpp tools. Readers get a reproducible evaluation methodology and practical guidance on working with two-bit quantization.








