模型

JinaAI 发布 jina-ocr-v1,在 25 种语言和低预算 GPU 上优化

Post-trained on DeepSeek-OCR, we strengthened multilingual support across 25 languages and optimized...

精选理由

JinaAI 推出了新OCR模型,支持25种语言,还能在nvidia L4这类低预算GPU上运行,性能还很强。

JinaAI 基于DeepSeek-OCR进行微调,推出新版本jina-ocr-v1。该模型在OmniDocBench v1.6上得分91.14,在olmOCR-Bench上得分83.4,在所有竞争者中页面吞吐量最高。

原文 · Jina AI

Post-trained on DeepSeek-OCR, we strengthened multilingual support across 25 languages and optimized...

Post-trained on DeepSeek-OCR, we strengthened multilingual support across 25 languages and optimized it for low-budget GPUs such as the nvidia L4. jina-ocr-v1 scores 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, and hits the highest page throughput among all competitors. 💬 1 🔄 0 ❤️ 0 👀 85 📊 1 ⚡