Sage 2 extraction model tops Extract Bench with 77.8% field accuracy, beating Cloudflare's Clairic
iamrobotbear · x · 2026-11-03
Developer omarsar0 highlights Sage 2, a new "system-one" extraction model from Levanto: you describe the fields you need and it returns structured values with exact source text and a probability for each result, complementing decision models.
Per Extract Bench v0.1 (mean of 10 datasets, held-out half), Sage 2 leads with 77.8% field accuracy and 72.2% on correctly answering "not in the document", ahead of NuExtract3 (76.1/57.0), Cloudflare's Clairic (75.3/59.0), Schematron-3B (59.9/54.3) and GLiNER2.5 (35.7/44.1). The author sees strong potential for processing and analyzing research papers.
More from Models
- fchollet: LLM writing's worst habit is turning every sentence into a punchline — fchollet · 2026-11-03
- Power user burns through $1,200/year Grok plan weekly, mulls $2,000 tier upgrade — eter_inquirer · 2026-11-03
- User on xAI's $1,200/yr Grok plan still hits 100% weekly usage cap — eter_inquirer · 2026-11-03
- DeepSeek's releases may be the year's most important: run a whole business on <$50/month — EXM7777 · 2026-11-03
- MIT shows LLMs can natively edit weights: 0.016% change lifts MATH-500 from 7.9% to 48.1% — lateinteraction · 2026-11-03
- OpenAI will soon release the model that solved a Millennium Prize Problem, leaker claims — kimmonismus · 2026-11-03