Sakhanda Wire
NVDA $153.61 -1.62% MSFT $506.27 -2.26% GOOGL $280.41 -1.17% META $623.08 -0.53% AMZN $257.66 -0.03%
← Back to the news

Liquid AI releases speculative decoding model for LFM2.5-VL-3B

Liquid AI has released LFM2.5-VL-3B-DSpark, an experimental draft model designed to accelerate generation by the LFM2.5-VL-3B vision-language model. The drafter has 279.5M parameters, proposes several tokens in advance, and lets the larger model verify them in a single pass. Liquid AI reports decoding gains of up to 3.13x on an Apple M5 Max and 2.66x on an NVIDIA H100.

The release is available on Hugging Face in Safetensors and GGUF formats, with support in SGLang, MLX-VLM and llama.cpp. Testing covered six vision-language task categories, using batch size 1, temperature 0 and 16-bit weights for the vision encoder and backbone. Under greedy decoding, the output matches the base model. With sampling, the target distribution is preserved, although higher temperatures can reduce token acceptance and throughput.

End-to-end improvements are smaller because image encoding and prompt processing are not accelerated. The drafter adds 8.9% to the deployed parameter count. The release uses the LFM Open License v1.0, which permits free commercial use only for companies with less than $10M in annual revenue. Quantized-model acceleration is not included.

Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.

Originally published by MarkTechPost on

Read the original on MarkTechPost ↗

The reporting and the image belong to MarkTechPost and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news