Qwen3.5 9B

unsloth/Qwen3.5-9B-GGUF:UD-Q8_K_XL

llama.cpp1 build benchmarkedbest build unsloth · Q8_K_XL32.7 GB24 tok/s

best build

Scores at a glance

measured avg35

Tool Calling497/27

Reasoning & Math216/28

0255075100

0–100 per task · n/m = cases passed (scores credit partial passes) · measured average covers 2/9 suite tasks

best build

config

deployment

run this buildllama.cpp

llama-server -hf unsloth/Qwen3.5-9B-GGUF:UD-Q8_K_XL -c 262144

Pulls the GGUF from Hugging Face and serves an OpenAI-compatible endpoint on :8080.

record

per-test results

Per-test scores, full transcripts, and the reproduce recipe.