Event ·Official first-hand

Jina-OCR-v1发布：低成本GPU文档解析模型

Jina-OCR-v1是一款端到端文档解析模型，专为在低成本GPU上运行设计。该模型结合了DeepSeek-OCR的压缩视觉编码器和30亿参数的专家混合解码器，使用FastMTP推测解码头实现每秒2.57页的处理速度。在OmniDocBench v1.6基准测试中得分为91.14，在olmOCR-Bench中得分为83.4。在NVIDIA L4等低成本GPU上，FastMTP比贪心自回归解码速度快一倍。

Current read

Jina团队发布了Jina-OCR-v1，能在低成本GPU上高效处理文档，速度翻倍且精度高。

3 sources73° AI heatLast updated 2026/9/2 21:49:21

Evidence

3 sources

Primary sourcearXiv: DeepSeek

2026/9/2 21:49:21View source

Related source 1DeepLearning.AI

2026/9/1 18:46:34View source

Related source 2NVIDIA AI

2026/9/4 16:47:25View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story