qwen
Qwen: Qwen2.5 VL 72B Instruct
qwen/qwen2.5-vl-72b-instruct
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
qwen/qwen2.5-vl-72b-instruct is the instruction-tuned 72-billion-parameter member of the Qwen2.5-VL vision-language family, listed on OpenRouter for text and image input producing text, with a 128,000-token context and a June 2024 knowledge cutoff. Its model card describes object and document analysis, chart, icon and layout understanding, visual grounding with bounding boxes or points, structured JSON outputs for forms and invoices, long-video comprehension, and visual-agent computer and phone use. Vendor-supplied benchmark tables look competitive but are not independently verified in this source pack.[1][2]
Strengths
- The model card reports strong document and OCR understanding, including analysis of text, charts, icons and layouts within images, plus structured JSON extraction for invoices, forms and tables.[2]
- Vendor documentation describes visual grounding via bounding boxes and points, agent-style computer and phone control, and comprehension of videos over one hour with event pinpointing.[2]
- The OpenRouter listing offers a 128,000-token context and structured-output parameters; the card also notes configurable image resolution to trade quality against compute.[1][2]
Limitations
- Evidence is a marketplace catalog plus the vendor's own card; benchmark figures are vendor-reported and unverified here, so validate on your own workloads.[1][2]
- OpenRouter lists only text and image input for this entry while the card claims video understanding, so video support through that endpoint is not confirmed.[1][2]
- Knowledge cutoff is 2024-06-30; the card requires building transformers from source and warns that video URL support depends on third-party backends.[1][2]
Where to use it
- Suggested for document intelligence: extracting structured data from invoices, forms, charts and scanned layouts.[2]
- Suggested for visual-grounding or UI-agent prototypes where bounding boxes, points or screen understanding matter.[2]
- Suggested for long-video question answering or event localization where the deployment actually accepts video input.[2]
Sources
- [1] OpenRouter catalog · Qwen: Qwen2.5 VL 72B InstructSources checked: Nov 13, 2026
- [2] Model documentation · Qwen: Qwen2.5 VL 72B InstructSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Qwen2.5-VL is proficient in recognizing common objects such as flowers, birds, fish, and insects. It is also highly capable of analyzing texts, charts, icons, graphics, and layouts within images.
Specifications
- Model ID
- qwen/qwen2.5-vl-72b-instruct
- Creator
- qwen
- Added to OpenRouter
- 01.02.2025
- Context length
- 128,000
- Modality
- text+image->text
- Input modalities
- text, image
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.800 / out $1
- Knowledge cutoff
- 30.06.2024
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | — |
| Coding | No measurement | — |
| Agents | No measurement | parameters: structured_outputs |
| Writing | No measurement | output: text |
| Reasoning | No measurement | — |
| Document analysis | No measurement | — |
| Vision | No measurement | input: image |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
No evaluations are supplied for this model.
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| Parasailfp8 | 128,000 | $0.800 | $1 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.