Event ·Multi-source

ExtractBench测试20+模型文档提取

ExtractBench测试20多个模型在文档提取任务上的表现，Qwen 3.8模型表现最佳，GLM-5.3-flash和qwen 3.8 flash新版本也刚刚发布。测试结果基于统一F1的“价值准确性”指标，不包含视觉定位和置信度评分。

Current read

ExtractBench最新测试了20多个模型，Qwen 3.8模型表现抢眼，新版本GLM-5.3-flash和qwen 3.8 flash也值得关注。

11 sources53° AI heatLast updated 2026/8/26 17:21:40

Evidence

11 sources

Primary sourceJerry Liu

2026/8/26 17:21:40View source

Related source 1Geek

2026/8/26 10:46:04View source

Related source 2OpenRouter

2026/8/26 14:14:48View source

Related source 3IT之家

2026/8/26 14:21:30View source

Related source 4ollama

2026/8/26 14:28:57View source

Related source 5Paul Couvert

2026/8/26 14:32:40View source

Related source 6lmarena.ai

2026/8/26 14:37:38View source

Related source 7elvis

2026/8/26 14:42:48View source

Related source 8shao__meng

2026/8/26 14:46:59View source

Related source 9The Rundown AI

2026/8/26 14:52:29View source

Related source 10量子位

2026/8/26 16:48:12View 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