Benchmark result
Spark-X2.6-4B-Q4_K_M on NVIDIA GeForce RTX 3060 — 72.1 tok/s
Measured with llama.cpp b11716-797de63b2 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
1st fastest of 4 runs of this model on NVIDIA GeForce RTX 3060 · median 68.3 tok/s.
sort
- 172.1tok/s54.5llama.cpp · Q4_K_Mthis run
- 268.5tok/s56.4llama.cpp · Q4_K_M
- 368.1tok/s67.3llama.cpp · Q4_K_M
- 467.2tok/s66.0llama.cpp · Q4_K_M
Reproduce this run
toolllama.cpp b11716-797de63b2
modelSpark-X2.6-4B-Q4_K_M (Q4_K_M)
context8,192
thinkingoff
samplingtemperature 1 · top_p 0.949999988079071 · top_k -1 · min_p 0.05000000074505806 · presence_penalty 0 · repeat_penalty 1
clientv0.4.96+102
Set those in llama.cpp, then:
llm-benchmark benchmark --model "Spark-X2.6-4B-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.