MobileLLM-125M
Meta's 125M deep-and-thin architecture for on-device LLMs.
About
MobileLLM-125M is Meta's sub-billion-parameter language model introduced in the February 2024 paper on optimizing LLMs for on-device use, later presented at ICML 2024. Its key finding was that deep-and-thin architectures with embedding sharing and grouped-query attention beat wider designs at this scale, giving a 2.7 percent accuracy boost over prior 125M state of the art on zero-shot reasoning.
The model family, including 125M, 350M, and larger variants, was released on Hugging Face in October 2024. Check the model card for current license terms; the weights are published for research and on-device experimentation.
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 — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 126; Trending score 74/100 from researcher assessment.