← Back to Model Beat
Models·3d ago·all news from September 18, 2026

Jina AI Releases jina-ocr-v1: A 3.4B MoE Document Parser With Built-In Speculative Decoding for Low-Budget GPUs

Jina AI has launched jina-ocr-v1, a 3.4B parameter document parser capable of converting PDFs, tables, and charts into Markdown. By utilizing a Mixture-of-Experts architecture and a built-in speculative decoding head, the model is designed to operate efficiently on lower-budget hardware.

Covered by 1 source

Related stories

ModelsGoogle Joins OpenAI, Anthropic, Meta in Disclosing AI HacksSep 18 · 21 sourcesModelsMozilla's new Smart Window assistant runs on Mistral's modelsSep 16 · 22 sourcesModelsSecurity researchers used Anthropic's Claude to hack OpenAI's internal systems in under 72 hoursSep 18 · 9 sourcesModelsAlibaba's open-weight Qwen-Image-2.1 claims to beat closed models in image generation with just 7 billion parametersSep 20 · 12 sources