LlamaIndex ExtractBench: Qwen 3.8 Leads Document Extraction Benchmark
NielsRogge · x · 2026-08-27
LlamaIndex released ExtractBench, a schema-guided extraction benchmark evaluating 20+ open-weight models across 8 domains and 67 document types.
Key Results:
- Qwen/Qwen3.8-27B leads with a score of 89.75.
- Earlier Qwen generations also show strong performance.
- Kimi-K3 ranks as the next best.
Notes:
- Results measure "value accuracy" via unified F1.
- Visual grounding and confidence scores are excluded; specialized OCR tools (like LlamaParse) show value when these metrics are included.
- Results for GLM-5.3-flash and Qwen 3.8 flash are pending.
Related event: LlamaIndex Launches ExtractBench; Qwen Tops Document Extraction(3 posts)→
More from Models
- Nathan Lambert slams Meta's Muse Open License as 'trash' — natolambert · 2026-10-16
- Google launches Gemini 4 Argon to Ultra and Pro subscribers, touting physics and 3D skills — ShikharMurty · 2026-10-16
- Google gives X users early access to test Argon, not just companies — lyraxana · 2026-10-16
- Grok Voice 3.0 Tops Design Arena Speech Leaderboard, Up 7.4 Points — XFreeze · 2026-10-16
- Robotics LLMs tested on 16 tasks: GPT-6 Astra hits 17.2% success vs Fable's 9.1% — pabbeel · 2026-10-16
- Meta open-sources Muse Spark: top US open model, scores just below DeepSeek — bindureddy · 2026-10-16