Two new MoE language models and a co-designed inference engine claim 20+ tokens/s CPU decoding under 1-8 GB memory with benchmark scores comparable to much larger models.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SmallThinker: A Family of Efficient Large Language Models Natively Trained for Local Deployment
Two new MoE language models and a co-designed inference engine claim 20+ tokens/s CPU decoding under 1-8 GB memory with benchmark scores comparable to much larger models.