Benchmark result
gemma-4-E4B-it-Q4_0 on NVIDIA GeForce RTX 4060 Laptop GPU — 99.9 tok/s
Measured with llama.cpp b11985-6d194531e on October 29, 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 11 runs of this model on NVIDIA GeForce RTX 4060 Laptop GPU · median 89.2 tok/s.
sort
- 199.9tok/s67.9llama.cpp · Q4_0this run
- 299.6tok/s67.2llama.cpp · Q4_0
- 398.5tok/s68.7llama.cpp · Q4_0
- 495.8tok/s68.9llama.cpp · Q4_0
- 595.2tok/s62.0llama.cpp · Q4_0
- 689.2tok/s72.8llama.cpp · Q4_0
- 785.7tok/s69.7llama.cpp · Q4_0
- 885.6tok/s70.1llama.cpp · Q4_0
- 985.3tok/s65.9llama.cpp · Q4_0
- 1084.8tok/s56.4llama.cpp · Q4_0
- 1154.1tok/s63.4llama.cpp · Q4_0
Reproduce this run
toolllama.cpp b11985-6d194531e
modelgemma-4-E4B-it-Q4_0 (Q4_0)
context32,768
think limitnone
samplingtemperature 1 · top_p 0.949999988079071 · top_k 64 · min_p 0.05000000074505806 · presence_penalty 0 · repeat_penalty 1
clientv0.4.98+102
Set those in llama.cpp, then:
llm-benchmark benchmark --model "gemma-4-E4B-it-Q4_0" --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.