Benchmark result

Qwen3.8-27B-i1-IQ4_XS-GGUF-Smaller on 2× NVIDIA GeForce RTX 5060 — 40.1 tok/s

Measured with llama.cpp b10971-96ffdc41c on September 15, 2026.

What the model built

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.

Reproduce this run

toolllama.cpp b10971-96ffdc41c

modelQwen3.8-27B-i1-IQ4_XS-GGUF-Smaller

context65,536

thinkingon

effortxhigh (model default)

samplingtemperature 0.20000000298023224 · top_p 0.949999988079071 · top_k 20 · min_p 0.05000000074505806 · presence_penalty 0 · repeat_penalty 1

clientv0.4.59+97

Set those in llama.cpp, then:

llm-benchmark benchmark --model "Qwen3.8-27B-i1-IQ4_XS-GGUF-Smaller" --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.