Benchmark result
Qwen3.8-27B-UD-Q8_K_XL on NVIDIA RTX A6000 — 20.7 tok/s
Measured with llama.cpp manual on October 28, 2026.
What the model built
Open full screen →The coding scenario asks for a playable game in a single HTML file. This is exactly what the model returned, unedited; the frame only adds a line that tells this page how tall it is.
How this run compares
1st fastest of 2 runs of this model on NVIDIA RTX A6000 · median 20.6 tok/s.
sort
- 120.7tok/s77.6llama.cpp · Q8_Kthis run
- 220.5tok/s84.3llama.cpp · Q8_K
Reproduce this run
toolllama.cpp manual
modelQwen3.8-27B-UD-Q8_K_XL (Q8_K)
context262,144
thinkingon
think limitnone
clientv0.4.98+102
Set those in llama.cpp, then:
llm-benchmark benchmark --model "Qwen3.8-27B-UD-Q8_K_XL" --tool "llama.cpp"The client prompts for context length and thinking mode, and for the KV cache dtype on Unsloth Studio. Sampling is left at the model default — the values above are what the tool reported using, not overrides the client sent.