Abra: Scaling Diffusion Image Training

arXiv:2608.17286v1 Announce Type: new Abstract: Compute-optimal scaling laws guide the training of frontier language models yet remain largely unexplored for visual generation. We present a systematic scaling law study for text-to-image diffusion models using Abra, a controlled family of flow-matching transformers trained across three orders of magnitude worth of compute (10^{19} to 10^{22} FLOPs), reaching significantly larger compute budgets than previous works. We demonstrate that diffusion models scale just as predictably as language models but require far more data to train optimally: compute optimality occurs at approximately 200 image tokens per parameter, ten times the Chinchilla compute-optimal prescription for LLMs. We show that unlike language models, diffusion models are robust to overtraining and that practitioners should err on the side of more data rather than a larger model. Finally, we show that this predictability extends beyond training loss to generative quality metrics, optimal CFG settings, representation quality, and even the shape of the training curves, which collapse onto a universal form.

Related

• FRIGID: Scaling Diffusion-Based Molecular Generation from Mass Spectra at Training and Inference Time
• Amortized Moment Matching for Visual Generation
• Tracking High-order Evolutions via Cascading Low-rank Fitting

Source: arXiv cs.LG | 2026-08-19