Summary:
- Liquid AI has introduced LFM2.5-DSpark, a new series of draft models designed to significantly accelerate the decoding speed of large language models.
- These models achieve up to 3.18x faster decoding performance, optimizing computational efficiency for real-time AI applications.