List your company

Large Language Model Inference on Llama 3.2 1B

1.94TPOTHBaseline
0.33840.75421.171.5858Mar 15, 2026

Evaluation results

MethodLinks
BaselineBackbone=Llama-3.2-1B2026.031.947.693.6
Vocab. TrimmingBackbone=Llama-3.2-1B2026.031.076.822.73
SVDSoftmaxBackbone=Llama-3.2-1B2026.030.616.362.27
FlashHeadBackbone=Llama-3.2-1B2026.030.46.152.06