SmolLM2
Family of tiny on-device language models in 135M, 360M, and 1.7B sizes.
About
SmolLM2 is Hugging Face's family of compact language models, released in February 2025 in 135M, 360M, and 1.7B parameter sizes. The 1.7B variant was trained on 11 trillion tokens of FineWeb-Edu, DCLM, the Stack, and curated math and code data.
Instruct versions were built with supervised fine-tuning plus direct preference optimization, and the 1.7B supports text rewriting, summarization, and function calling. The family is a reference point for capable on-device inference, with Apple shipping Core AI export support for all three sizes. Release timing below is era-level.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 110; Trending score 82/100 from researcher assessment.