Benchmark result
Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp on AMD Radeon RX 9070/9070 XT/9070 GRE — 32.0 tok/s
Measured with llama.cpp b11026-b49650adb on September 18, 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
2nd fastest of 2 runs of this model on AMD Radeon RX 9070/9070 XT/9070 GRE · median 45.5 tok/s.
sort
- 159.1tok/s78.8llama.cpp
- 232.0tok/s81.5llama.cppthis run
Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp on other hardware
Reproduce this run
toolllama.cpp b11026-b49650adb
modelQwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp
context128,144
effortxhigh (model default)
samplingtemperature 1 · top_p 0.949999988079071 · top_k 20 · min_p 0 · presence_penalty 0 · repeat_penalty 1
clientv0.4.59+97
Set those in llama.cpp, then:
llm-benchmark benchmark --model "Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp" --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.