FS-DFM: New Apple AI Models Output Long Texts Very Quickly
With "Few-Step Discrete Flow-Matching," Apple wants to put the turbo on the output of large language models. A US university is a cooperation partner.
Large language models generate texts – something Apple now claims to have accelerated again.
(Image: Photo Kozyr/Shutterstock.com)
Apple's AI research department has introduced a new approach to accelerating text generators. Using so-called Few-Step Discrete Flow-Matching Language Models, or FS-DFM for short, the throughput rate ("faster sampling and corresponding latency/throughput gains") can be increased up to 128-fold, measured by the output of 1024 tokens. FS-DFM combines 1024 individual steps of conventional models into just eight steps, the scientists explain.
Technology from the field of image generation
In the paper, published on the preprint server arXiv, it is further stated that FS-DFM should generate long texts particularly accurately. This uses diffusion technology, a generation method that was originally used primarily for image generation but is now also used with Large Language Models (LLMs) because it promises efficiency gains. Competing systems such as OpenAI's GPT-5, on the other hand, rely on transformer technology, which uses so-called autoregression.
Videos by heise
Diffusion models have the fundamental advantage that they can generate multiple output tokens, i.e., the individual components of the text output, simultaneously. They are then further refined until the output is sent to the user. Flow-matching models are intended to further accelerate this, which Apple has now succeeded in doing – whereby existing methods are said to be outdated.
GPT-based LLMs are still mainstream
One problem remains that the selection of publicly available discrete flow-matching models is still small – the AI scene continues to rely primarily on the GPT-based approach for text generators. Apple's researchers want to help here by making their own DFM and FS-DFM models publicly available. Details can be found in the paper.
The project was created in collaboration with researchers at Ohio State University. Apple's AI research department had previously experimented with flow-matching models in other areas, including for programming and for protein folding. The company hopes for breakthroughs for its products as a result. Apple still does not officially offer its own chatbot.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
(bsc)
