Hugging Face has released SmolLM2 today, a set of new compact language models that achieve impressive performance while requiring significantly less computational resources compared to larger models. The new models are released under the Apache 2.0 license and come in three sizes—135M, 360M, and 1.7B parameters—suitable for deployment on smartphones and other edge devices with limited processing power and memory.
The SmolLM2-1B model outperforms Meta's Llama1B model in several key benchmarks, particularly excelling in scientific reasoning and common sense tasks. The model performs better than most large competitive models on cognitive benchmarks, using a diverse combination of datasets including FineWeb-Edu and specialized math and coding datasets.
The release of SmolLM2 comes at a critical time when the AI industry is struggling with the computational demands of running large language models (LLMs). While companies like OpenAI and Anthropic continue to push the boundaries of model size, there is an increasing recognition of the need for efficient, lightweight AI that can run locally on devices.
SmolLM2 offers a different approach, bringing powerful AI capabilities directly to personal devices, pointing towards a future where more users and companies can access advanced AI tools, not just tech giants with massive data centers. These models support a range of applications, including text rewriting, summarization, and function calls, suitable for deployment in scenarios where privacy, latency, or connectivity constraints make cloud-based AI solutions impractical.
While these smaller models still have limitations, they represent part of a broader trend towards more efficient AI models. The release of SmolLM2 indicates that the future of AI may not only belong to increasingly larger models but also to more efficient architectures that provide powerful performance with fewer resources.