PaddlePaddle/PaddleOCR-VL-1.5
Primitive: /extract · Extract ·
PaddleOCR-VL
PaddleOCR-VL-1.5: Towards a Multi-Task 0.9B VLM for Robust In-the-Wild Document Parsing
MultimodalMultilingualLong contextEntities
Overview
Hardware: — drives latency, throughput & cost
| Size | 959M params |
|---|---|
| Tasks | /extract |
| License | apache-2.0 |
| Languages | en, zh, multilingual |
| Latency | 12.4 s |
| Throughput | 989 tok/s |
| Cost | $0.225 /1M tok |
Cost is approximate — computed from list GPU prices; your actual price depends on the provider you deploy SIE with.
Extraction
| Output kinds | Entities |
|---|---|
| Inputs | image |
| Max sequence length | 16,384 |
Benchmarks
olmOCR-Bench
Document text extraction accuracy across arxiv math, old scans, multi-column layouts, and tables
Corpus: 1,403 Queries: 1,403
default
Quality
accuracy 0.5172
Performance L4 b1 c4
Extract 0.5 mpix/s
Extract p50 12.4s
sglang
Quality
accuracy 0.5179
default_limit-50
Quality
accuracy 0.4886
Compare (0)Compare →