Accelerating vision-language models with LFM2.5-VL-DSpark
Covered by 2 sources
Read the full articleA new DSpark draft model for the vision-language model LFM2.5-VL-3B has been released, offering up to 3.13x faster decoding speeds with minimal memory increase. It supports llama.cpp, MLX-VLM, and SGLang integrations for efficient vision-language tasks.
