Document layer for enterprise AI that turns PDFs, images, and spreadsheets into LLM-ready JSON and Markdown
Generative video models for animation
Stage
Not given
Not given
Founded
Not given
Not given
Based in
US
US
Pricing model
Not given
Not given
Pricing
Free/Hobby: $0, up to 2,000 pages included, $0.015 per additional page. Standard: $250/month, includes up to 20,000 pages, $0.012 per additional page. Enterprise: Custom per-page rate with volume-based pricing.
Not given
Key features
Parse documents to LLM-ready Markdown with vision models that preserve structure
Extract fields into JSON with domain-awareness and confidence scores
Split multi-document files with parent-child indexing
Attention-guided heatmaps for reading pages like humans
Dual-stream vision model processing data and layout in parallel
Domain-specific decoder trained on millions of enterprise documents
Supports multiple file types: PDFs, images, Word, Excel, PowerPoint, scans
Deployment options: managed cloud, VPC, on-premises, or air-gapped
REST API and Studio dashboard for integration
Schema-validated outputs with cross-field constraints
Pyramid-JiT (P-JiT) architecture with 32×32 token reduction
4× more aggressive token reduction than Linum v2
Converges in 11× fewer samples than Linum v2
2B-parameter model with frozen Qwen text encoder
Pixel-space diffusion with multi-resolution readouts
PixelREPA alignment with DINOv3-L
Curriculum learning over noise schedule
What makes it different
#1 ranked accuracy on olmOCR-Bench against leading frontier labs and IDP vendors
Only solution that reliably handled complex tables and nested formats where others failed
Simple per-page pricing with no credit-based ambiguity
Preserves document structure and hierarchy unlike traditional OCR
Low-cost training: full 2B model trained for just $21K in compute
Controls baked into model training rather than post-hoc fine-tuning
Pixel-space compression without separate VAE bottleneck
Apache 2.0 open release with full model weights and code
Unsiloed AI
Linum