Liquid AI released two open-weight bidirectional encoders, LFM2.5-Encoder-230M and LFM2.5-Encoder-350M, built on the LFM2 hybrid backbone with 8,192-token context windows. (Liquid AI) The 350M model ranked fourth among 14 models on a 17-task benchmark suite covering GLUE, SuperGLUE, and multilingual evaluations, trailing only larger models. (Liquid AI)
Fireworks AI released Fireworks Nexus, a routing and cost-control layer designed to integrate with existing developer tooling and direct routine coding work to open-weight models. (Fireworks AI) The platform targets engineering organizations seeking to reduce API costs for repetitive tasks. (Fireworks AI)