LandingAI has released Agentic Document Extraction, or ADE, Gen2, a rebuilt version of its document intelligence stack built around the new DPT-3 model family. The company said the release is now generally available, and developers can start using it for free in the ADE sandbox.
DPT-3 models sit at the center of the new stack
According to LandingAI, Gen2 treats documents as tree structures. Its billing model now charges by returned character count rather than by page, and each answer can be traced back to a specific line or word on the page.
Gen2 splits parsing into two models. DPT-3 Verity is aimed at deterministic transcription for digitally created documents and returns bounding boxes and confidence scores for each word. DPT-3 Pro reads page layout first, then detects more complex elements including tables, graphics, handwriting, non-Latin characters, and LaTeX mathematical formulas.
Verity is priced below Pro, with auto-routing planned for fall 2026
LandingAI said DPT-3 Verity is priced at about 40% of DPT-3 Pro. The company also said it plans to introduce automatic routing between the two models in fall 2026.
Pricing shifts from fixed per-page fees to component plus character charges
The pricing structure has changed as part of the update. Under DPT-2, pricing was a flat 3 credits per page. Under DPT-3, charges are calculated as the sum of page components and output-character components.
For example, priority-tier DPT-3 Pro is priced at 1 credit per page plus 0.5 credits per 1,000 characters. LandingAI said it expects costs for mixed workloads to fall by 25% to 80%, and it claimed that parsing with the Verity standard tier can cost less than one cent per page.