OSS
Apple released LensVLM-9B weights, a VLM that reads long text as compressed images and expands only the pages it needs
The weights went up on Hugging Face on 2026-09-21, with 160 likes so far, and the HN post reached 79 points. The model is built on Qwen3.5-9B-Base and follows the May paper arXiv:2605.07019. It renders context as images at 5x, 10x or 15x compression and calls learned tools to decompress the pages that matter. Apple reports accuracy on par with full text at 4.3x effective compression, and says it beats retrieval and compression baselines up to 10.1x on seven QA benchmarks. The license is Apple's research model license, which rules out commercial use.
Source
↳ Follow the thread