Liquid AI has introduced LFM2.5-VL-3B-DSpark, an experimental speculative decoding draft model built for its LFM2.5-VL-3B vision-language model. The company said the draft model adds about 280 million parameters and is designed to speed up decoding without changing model outputs. According to the Liquid AI team, decoding throughput improved by as much as 3.13x on Apple silicon and up to 2.66x on NVIDIA H100 hardware. The model has already been deployed, and its weights are available on Hugging Face in both Safetensors and GGUF formats. It also received day-one support across SGLang, MLX-VLM, and llama.cpp. Liquid AI labeled the release as experimental and said it is distributed under the LFM Open License v1.0, which allows free commercial use for companies with annual revenue below $10 million. The update was cited by Techub, referencing MarkTechPost.
Liquid AI has released LFM2.5-VL-3B-DSpark, an experimental speculative decoding draft model for its LFM2.5-VL-3B vision-language model.
The company said the draft model adds roughly 280 million parameters and can accelerate decoding without changing the model's output. The Liquid AI team reported decoding speed gains of up to 3.13x on Apple silicon and up to 2.66x on NVIDIA H100.
The model is already deployed. Its weights are available on Hugging Face in Safetensors and GGUF formats, with day-one support provided in SGLang, MLX-VLM, and llama.cpp.
Liquid AI labeled the release as experimental and said it is offered under the LFM Open License v1.0. Under that license, companies with annual revenue below $10 million can use it commercially for free.
Techub cited MarkTechPost as the source of the report.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
