Apple's DFN5B-CLIP Hits 84% ImageNet Accuracy by Teaching AI to Pick Its Own Data

Apple's DFN5B-CLIP-ViT-H-14-378 keeps climbing Hugging Face charts, hitting 84.2% zero-shot ImageNet accuracy by rethinking how training data is filtered.

Read the full story at AlphaSignal ↗

Timeline · 1 report

2026-10-17 16:00 · AlphaSignal
Apple's DFN5B-CLIP Hits 84% ImageNet Accuracy by Teaching AI to Pick Its Own Data

More stories

Meta releases Muse Spark Open, an 816B-parameter AI model optimized for Mac Studios — Vercel Blog
Apple tightens App Store rules to block AI nudify apps — Engadget
Microsoft Surface Laptop Ultra features Nvidia RTX Spark for local AI — Tom's Hardware
Hugging Face details 2.8 trillion token Indic pretraining pipeline — Hugging Face Blog
Hugging Face releases Beam, a 501B parameter open model — Hugging Face Blog
Granite Switch 4.1 offers chat, decisions, and speech in one endpoint — Hugging Face Blog
Apple details local and distributed LLM inference using MLX and RDMA on Mac Studio — Apple Machine Learning Research
Apple Research Proposes Selection-Based Structured Reasoning for Multimodal Agents — Apple Machine Learning Research

Get the daily brief of stories like this at 6:30 every morning →