Key facts
| Vision input | Image understanding runs on the OpenAI-compatible chat completions endpoint with supported multimodal models |
| Example model | plugsky-gemma-4 handles text plus image inputs such as screenshots, forms and product photos |
| Model access | 30+ models from efficient to frontier tiers behind one key |
| Dedicated media endpoints | Image generation and editing endpoints are labelled coming soon |
| Deployment | Hosted, VPC, on-prem and air-gapped options on Enterprise |
| Data handling | Data encrypted at rest and in transit; API data is not used to train models |
| Audit | Per-request logs with model, tokens, latency and user; SIEM export; retention up to 7 years |
| Free plan | plugsky-micro and plugsky-lite, no card; 14-day full-access trial |
TL;DR
- Turn document images into structured fields your workflow can validate.
- JSON mode and function calling keep extraction typed and testable.
- Region pinning, BYOK and audit logs keep the pipeline reviewable.
- Set a confidence threshold and route the rest to human review.
- Pilot one document type and measure field accuracy before scaling.
How it works, step by step
- Pick one high-volume document type with a stable format.
- Label a sample set and define the fields that must be extracted exactly.
- Send images through the OpenAI-compatible chat completion with a JSON schema.
- Measure field-level accuracy and set the human-review threshold.
- Pin the region and enable audit logging before processing real documents.
- Move to VPC, on-prem or air-gapped deployment if the corpus requires it.
- Review model choice and thresholds monthly against exception rates.
Original data
Try it yourself
Open the model availability tracker →
Why banks need image understanding
Banking operations run on documents: identity papers at onboarding, cheques at clearing, statements in credit review, trade documents at settlement. Most of that work is still checked by people reading images, and the queue grows with volume. Vision models can turn an image into structured fields — name, date, amount, reference — and hand it to your existing workflow.
The goal is not a model that looks at a photo. It is a pipeline that extracts fields into a schema, flags low-confidence cases for human review, and keeps an audit trail of what was read, by which model, at what time. Image generation is a separate capability and is labelled coming soon, so design around understanding first.
How vision works on the Plugsky API
Image understanding does not need a new platform. Add the image to the message content of a standard chat completion, and the response arrives exactly like a text completion: streamable, validatable with JSON mode, and usable with function calling. plugsky-gemma-4 covers high-volume triage such as screenshots, forms and document images, while stronger models handle the hard cases. Because 30+ models sit behind one key, changing tier is a model-name change. For banks, the design goal is fields into a schema with a human threshold, not a model that merely looks at documents.
The controls banks should configure first
Controls follow the same pattern as text workloads. Pin processing to a region-locked plane, or move document extraction to a VPC, on-prem or air-gapped deployment when the corpus demands it. Keys stay with the bank through BYOK with AWS KMS, Azure Key Vault, HashiCorp Vault or an on-prem HSM, and per-request logs record model, tokens, latency, user and region for retention up to 7 years.
On the application side, constrain output with JSON mode or function calling so extracted fields land in a typed schema, and set a confidence threshold below which a human reviews the document. That keeps automation accountable and makes exceptions measurable.
A practical pilot path
Begin with one document type, high volume and low ambiguity — a statement format or a standard onboarding form. Build a labelled sample, measure field-level accuracy against the current manual process, and define an automatic-review threshold. Prototype on the free plan with plugsky-micro and plugsky-lite, then move to the required deployment plane for real documents. See the live pricing page for current plans.
Honest comparison
| Capability | Plugsky | Vision-only API vendor | Building in-house |
|---|---|---|---|
| Image understanding | Live on supported multimodal chat models | Usually the core product | You host and serve vision models |
| Image generation | Labelled coming soon | Often available today | Custom deployment work |
| API shape | OpenAI-compatible chat completions | Vendor-specific SDK | You standardise it yourself |
| Model choice | 30+ models behind one key | Narrower catalogue | One model per integration |
| Deployment | Hosted, VPC, on-prem and air-gapped | Usually hosted only | Your infrastructure only |
| Controls | Region pinning, BYOK, per-request audit logs, SIEM export | Varies by vendor | You build the pipeline |
Frequently asked questions
Is image understanding live on Plugsky?
Yes — image understanding runs on supported multimodal chat models. Dedicated image generation and editing endpoints are labelled coming soon, so plan them as roadmap items.
Does this replace our OCR stack?
Treat it as complementary. Vision models handle layout, context and ambiguous fields well; deterministic OCR still helps where text is clean and auditability demands exact character output.
Do we have to change our application code?
No. Image understanding uses the same OpenAI-compatible chat completions endpoint, so your SDK, auth and error handling carry over; the message content simply includes an image.
Is there a free plan?
Yes — plugsky-micro and plugsky-lite with 2 API keys and no credit card. A 14-day full-access trial is also available.
How is pricing structured?
Self-serve plans are flat monthly with unlimited fair-use usage and no per-token charges or overage fees. See the live pricing page for current plans.
Is image generation available?
Not yet — image generation and editing endpoints are labelled coming soon. Image understanding is live today on supported multimodal chat models.