Add Gemma 3 1B, per-model accuracy table

Gemma 3 1B int8, [out, in] layout

SmolLM2 135M int8, [out, in] layout

Qwen3 0.6B int8, [out, in] layout

Add model card

initial commit