Apple AI Updates: October 9, 2026

1. Apple’s Normalizing Trajectory Models Generate Images in Four Steps With Exact Likelihood

Apple. Apple Machine Learning Research published Normalizing Trajectory Models (NTM) on October 8, a NeurIPS paper from Jiatao Gu, Tianrong Chen, Ying Shen, David Berthelot, Shuangfei Zhai, and Josh Susskind. NTM replaces the small Gaussian denoising steps used in diffusion models with conditional normalizing flows, combining shallow invertible blocks at each step with a deep parallel predictor across the whole trajectory, so the model keeps exact likelihood over the full generation path. It can be trained from scratch or initialized from pretrained flow-matching models. Because the likelihood is exact, the authors train a lightweight denoiser on the model’s own score function, and they report that NTM matches or beats strong text-to-image baselines with four sampling steps. The page does not mention a code release. Source