LFM2.5-VL-3B generates text faster locally: up to 3.13× on M5 Max

On Sep 24, Liquid AI introduced DSpark for LFM2.5-VL-3B: generation on M5 Max sped up by 2.30–3.13×. A full VLM response arrived 1.56–2.62× faster. The weights are already available on Hugging Face in Safetensors and GGUF.