Benchmark result
qwen2.5-0.5b-instruct-q4_k_m on AMD Radeon RX 7900 XTX — 258.7 tok/s
Measured with llama.cpp manual on October 20, 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
4th fastest of 5 runs of this model on AMD Radeon RX 7900 XTX · median 261.4 tok/s.
sort
- 1341.5tok/s6.4llama.cpp · Q4_K_M
- 2340.9tok/s10.0llama.cpp · Q4_K_M
- 3261.4tok/s7.5llama.cpp · Q4_K_M
- 4258.7tok/s10.0llama.cpp · Q4_K_Mthis run
- 5243.5tok/s10.7llama.cpp · Q4_K_M
Reproduce this run
toolllama.cpp manual
modelqwen2.5-0.5b-instruct-q4_k_m (Q4_K_M)
context8,192
think limitnone
samplingtemperature 0.800000011920929 · top_p 0.949999988079071 · top_k 40 · min_p 0.05000000074505806 · presence_penalty 0 · repeat_penalty 1
clientv0.4.96+102
Set those in llama.cpp, then:
llm-benchmark benchmark --model "qwen2.5-0.5b-instruct-q4_k_m" --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.