1. Hugging Face launched SmolLM2, a set of compact language models designed to be efficient and powerful. 2. The models come in three sizes: 135M, 360M, and 1.7B parameters. 3. SmolLM2-1B outperforms Meta’s Llama1B on key benchmarks, especially in scientific reasoning and common sense. 4. These models are perfect for smartphones and edge devices, not just large data centers. 5. SmolLM2 signals a shift toward efficient AI that’s accessible to more users and companies.