LLM Inference on BitNet 2B Intel Core Ultra b1.58
556Memory (MB)Litespark-Inference
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Litespark-InferenceImplementation Kernel=AVX-VNNI, Instruction Set=AVX-VNNI (256-bit)2026.05 | 556 | 310 | 8.5 | 21.3 | |
| PyTorchImplementation Kernel=Baseline2026.05 | 7,750 | 2,580 | 0.4 | — |