# Apple — The Design Space of Tri-Modal Masked Diffusion Models

- Company: Apple (apple.com)
- Announced: 2026-10-14
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://machinelearning.apple.com/research/tri-modal-masked-diffusion
- Record: https://forck.live/items/22236-the-design-space-of-tri-modal-masked-diffusion-models
- Subject: Machine Learning Research

Apple researchers introduced the first tri-modal Masked Diffusion Model (MDM) pretrained from scratch on text, image-text, and audio-text data. They systematically analyzed multimodal scaling laws, modality mixing ratios, noise schedules, and batch-size effects, proposing a novel SDE-based reparameterization that decouples physical and logical batch sizes. A preliminary 3B-parameter model trained on 6.4T tokens achieved strong results in text generation, text-to-image, and text-to-speech tasks.

## Evidence

Verbatim from https://machinelearning.apple.com/research/tri-modal-masked-diffusion:

> Diverging from previous approaches, we introduce the first tri-modal Masked Diffusion Models (MDM) pretrained from scratch on text, image-text, and audio-text data.

---

Record: https://forck.live/items/22236-the-design-space-of-tri-modal-masked-diffusion-models
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md