Quail speeds up AI extraction up to 19x with in-document speculative drafting

Shreya Shankar and Charles Frye describe how Quail's AI.EXTRACT uses speculative decoding with draft tokens copied directly from the source document, achieving an average 4.7x and up to 19x speedup.

2026-11-10 ~ 2026-11-10 · 3 related posts

• Drafting from the document makes AI extraction 4.7x faster via adaptive speculative decoding — sh_reya · 2026-11-10
• Copy-from-document speculative decoding speeds AI extraction 4.7x (up to 19x) vs naive vLLM — sh_reya · 2026-11-10
• Quail's Draft-from-Document AI Extraction Is 4.7x Faster, Up to 19x — sh_reya · 2026-11-10