Tencent released Youtu-Parsing-Omni weights, inference examples and vendor benchmarks.
Tencent releases one parser for documents, charts, audio and video
2026-10-09Verified announcementTRIAL
Tencent released the 5B Youtu-Parsing-Omni weights, which parse documents, natural images, charts, geometry, audio and video into structured JSON. The model card includes Transformers and vLLM usage and reports OmniDocBench and OmniParsingBench results, while evaluation code is still marked TBD.
What this means for you
For multimodal ingestion pipelines, one model across input types could reduce routing and schema stitching. Benchmarks are vendor-reported, evaluation code is not yet public and the license is custom, so start with a small test on your own mixed inputs.
Evidence & related material
Technical details & scores
| Parameters | 5B |
|---|---|
| Inputs | documents, natural images, charts, geometry, audio and video |
| Output | structured JSON |
| Runtime | Transformers custom code and vLLM plugin examples |
| Vendor Results | OmniDocBench v1.6 96.96; OmniParsingBench 75.08 |
| Evaluation Code | TBD in model card |
| License | custom Youtu license |
Technical impact 4
Engineering value 4
Adoption signal 1
Maturity 2
Verification cost 3
Risk 3