Timeline of Llama-3.2-3B

Llama-3.2-3B scores 21.05% on τ²-bench

Llama-3.2-3B scores 26.19% on IFBench

Llama-3.2-3B scores 5.35% on Humanity's Last Exam

Llama-3.2-3B scores 25.45% on GPQA

Llama-3.2-3B scores 5.7 on Artificial Analysis Intelligence Index

Llama-3.2-3B: openness changed from restricted to open-weights

New model: Llama-3.2-3B (Meta AI)