Key facts
| Model class | Vision-language profile in the Plugsky catalogue |
| Best for | Screenshots, forms, product images and vision-triggered tool calls |
| Context class | 128K-class window; live limits are published per model |
| Capabilities | Vision inputs, streaming, function calling and JSON mode |
| Pricing tier | Paid-plan model; free plan covers plugsky-micro and plugsky-lite |
| API | OpenAI-compatible /v1/chat/completions — no SDK changes |
| Resilience | Backup upstream plus same-profile fallback peers |
| Product status | Live |
TL;DR
- Vision and text together, with function calling on top.
- Best for screenshots, forms and workflows that must act.
- Structured output via JSON mode suits extraction pipelines.
- It is image-input only — generation is a separate endpoint.
- Check /models for live vision flags and limits.
How it works, step by step
- Read the live model card at /models for vision flags and limits.
- Define the visual task: answer a question, extract fields or trigger an action.
- Send the image with clear text instructions and keep the prompt short.
- Use JSON mode plus a schema validator for extracted fields.
- Connect tools only where the visual decision must trigger an action.
- Score accuracy on a labelled image set, including rotated and low-quality scans.
Try it yourself
What plugsky-qwen-vl is
plugsky-qwen-vl is the vision-language profile in the Plugsky catalogue. It accepts images alongside text in the chat messages array and can respond with answers, structured data or tool calls. That combination — seeing an image and acting on it — is what separates it from caption-only vision models.
It is listed as 128K-class on the live model card, and like every profile, the engine serving it can change. Check /models for current capability flags, and the docs for accepted image formats.
When to choose it
Choose plugsky-qwen-vl when a workflow must both understand an image and do something with it: reading a screenshot and filing a ticket with the right fields, checking a product photo against a catalogue and calling an inventory tool, or extracting totals from a scanned invoice and writing them to a system of record.
For high-volume captioning or simple triage, a faster compact model may be enough. For deep reasoning over a complex diagram, consider escalating to a stronger vision-capable tier after testing both on your images.
Vision trade-offs
Images consume context, and quality depends heavily on what you send:
- Resize and crop before sending; resolution above what the task needs is wasted context.
- One relevant image beats several redundant ones in both cost and accuracy.
- Validate extracted fields in code — vision extraction is not a substitute for schema checks.
- Keep an image eval set with hard cases: rotation, glare, handwriting and UI clutter.
Self-serve plans are flat monthly with fair-use usage — see the live pricing page for current plans and limits.
How to switch to plugsky-qwen-vl
Switching is one model name, plus images in the messages array: {"model": "plugsky-qwen-vl", "messages": [{"role": "user", "content": [{"type": "text", "text": "Extract the invoice total."}, {"type": "image_url", "image_url": {"url": "..."}}]}]}.
Test on your own images before moving traffic, and keep a stronger vision-capable tier configured as the escalation path for images that need deep reasoning.
Honest comparison
| Dimension | plugsky-qwen-vl | Fast vision tier | Compact multimodal tiers |
|---|---|---|---|
| Best fit | Vision plus tool calling | High-volume visual triage | Simple extraction and captioning |
| Tool use | Yes | Limited | Limited |
| Structured output | JSON mode | Varies by model | JSON mode on some tiers |
| Context class | 128K-class | 128K-class | 128K-class |
| Use as | Vision workflows that act | First hop for volume | Specialist extraction |
| Plan | Paid | Paid | Paid |
Frequently asked questions
Is plugsky-qwen-vl free?
No — it is a paid-plan model. The free plan covers plugsky-micro and plugsky-lite with no card required, and a 14-day full-access trial lets you evaluate paid tiers first.
What is plugsky-qwen-vl best at?
Vision-language work that also needs actions: screenshots with tool calls, form and invoice extraction, product-image questions and visual pipelines that write results into other systems.
Can it generate images?
No — it accepts images as input. Image generation and editing are separate endpoints, so check the docs for their current status before designing around them.
How are images sent?
Inside the chat messages array alongside text, using the standard OpenAI-compatible multimodal shape. The docs cover accepted formats and size guidance.
What context window does it have?
It is 128K-class today, but the live window and output limit are published per model on the catalogue.
How do I evaluate vision quality?
Build a labelled image set per task, include hard cases such as rotation and glare, and measure field-level accuracy for extraction or rubric scores for open questions.
How is pricing structured?
Self-serve plans are flat monthly with fair-use usage rather than per-token billing. See the live pricing page for current plans.