Liquid AI Unveils LFM2.5-Encoder Models for Efficient Long-Context CPU Inference
Covered by 2 sources
Liquid AI has introduced the LFM2.5-Encoder-230M and 350M models, bidirectional encoders optimized for fast inference on CPUs with input contexts up to 8,192 tokens. These models, derived from LFM2.5 decoders, deliver strong benchmark performance and support various NLP applications including classification and search.