Benchmark result
LFM3-Omni-2B-Q8_0 on NVIDIA GeForce RTX 3060 — 136.6 tok/s
Measured with llama.cpp b12015-a34de8618 on November 2, 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
3rd fastest of 4 runs of this model on NVIDIA GeForce RTX 3060 · median 137.1 tok/s.
sort
- 1140.6tok/s29.3llama.cpp · Q8_0
- 2137.7tok/s25.8llama.cpp · Q8_0
- 3136.6tok/s52.4llama.cpp · Q8_0this run
- 4130.5tok/s40.6llama.cpp · Q8_0
Reproduce this run
toolllama.cpp b12015-a34de8618
modelLFM3-Omni-2B-Q8_0 (Q8_0)
context32,768
thinkingoff
samplingtemperature 0 · top_p 0.949999988079071 · top_k 1 · min_p 0.05000000074505806 · presence_penalty 0 · repeat_penalty 1.100000023841858
clientv0.4.98+102
Set those in llama.cpp, then:
llm-benchmark benchmark --model "LFM3-Omni-2B-Q8_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.