How to Run Underdog Saluki 27B: 2-Bit Qwen3.8-27B Tool-Calling Guide
MarkTechPost Sana Hassan
Underdog Saluki 27B was tested against the underlying Qwen3.8-27B GSQ-RCO 2-bit files by diffing their GGUF headers and then serving both on one GPU to measure tool-calling performance with the same BFCL checker. The evaluation ran on 120 seeded tasks from the Berkeley Function Calling Leaderboard. The result is that Saluki keeps tool calling at about 2 bits per weight and reports a higher score than the full 54 GB model in the notebook’s benchmark setup.
Why it matters
Explore a complete coding guide comparing Underdog Saluki 27B with ISTA's GSQ-RCO quantization, examining GGUF headers, memory usage, and BFCL tool-calling performance in stock llama.cpp The post How to Run Underdog Saluki 27B: 2-Bit Qwen3.8-27B Tool-Calling Guide appeared first on MarkTechPost.
Related stories
Meet the Underdog Saluki 27B: A 2-bit Qwen3.8-27B That Beats the Original at Tool Calling
MarkTechPost · 2 weeks ago ·
41
Qwen 3.8 27B scores 52 on the Artificial Analysis Intelligence Index
Simon Willison’s Weblog · 2 months ago ·
8
Qwen3.8 27B addition in words
Simon Willison’s Weblog · 2 weeks ago ·
14