Vision OCR Markdown

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit

Not collected

Maximum input tokens

Not collected

Maximum output

Not collected

Output consists of recognized text and coordinates; a generative token allowance is not specified.

Reasoning

Not collected

Tool calling

Not collected

Structured output

Not collected

Fixed OCR response fields are documented; this does not establish user-defined LLM JSON Schema support.

Streaming

Not collected

Batch processing

Not collected

Fine-tuning

Not collected

Knowledge cutoff

Not collected

Release date

2025-06-30Source for Release date (opens in a new tab)

Date these API models were added, not a weight-training date.

Parameters

Not collected

Languages

View all 37 entries

• Azerbaijani (az)
• Bashkir (ba)
• Belarusian (be)
• Bulgarian (bg)
• Bosnian (bs)
• Czech (cs)
• Chuvash (cv)
• Danish (da)
• German (de)
• English (en)
• Spanish (es)
• Estonian (et)
• Finnish (fi)
• French (fr)
• Hungarian (hu)
• Indonesian (id)
• Italian (it)
• Kazakh (kk)
• Kirghiz (ky)
• Lithuanian (lt)
• Latvian (lv)
• Maltese (mt)
• Dutch (nl)
• Norwegian (no)
• Polish (pl)
• Portuguese (pt)
• Romanian (ro)
• Russian (ru)
• Yakut (sah)
• Slovak (sk)
• Slovenian (sl)
• Serbian (sr)
• Swedish (sv)
• Tajik (tg)
• Turkish (tr)
• Tatar (tt)
• Uzbek (uz)

Source for Languages (opens in a new tab)

Supported language options, not simultaneous recognition of every language in one operation.

Context caching

Not collected

Input modalities

Image, PDFSource for Input modalities (opens in a new tab)

Output modalities

TextSource for Output modalities (opens in a new tab)

Input file formats

JPEG, PNG, PDFSource for Input file formats (opens in a new tab)

Image input limits

10 MB per file; 20 megapixels per image.Source for Image input limits (opens in a new tab)

Document limits

Up to 200 PDF pages using asynchronous recognition. Single-page PDFs can use the synchronous route.Source for Document limits (opens in a new tab)

Data retention

Asynchronous recognition results are retained for 3 days.Source for Data retention (opens in a new tab)

Result retention is not a universal no-training or no-logging commitment.

Native model options

markdownSource for Native model options (opens in a new tab)

Exact model parameter; languageCodes separately chooses the applicable language model.

Output formats

JSON, MarkdownSource for Output formats (opens in a new tab)

Billing basis

Each successfully recognized image or PDF page is a separate billable unit. Repeating recognition is separately billed.Source for Billing basis (opens in a new tab)

Language pack selectors

One language model per operation: a Latin-Cyrillic model, or Arabic, Greek, Hebrew, Armenian, Japanese, Georgian, Korean, Thai, Vietnamese or Chinese, each with English and Russian.Source for Language pack selectors (opens in a new tab)

Usage metering

Not collected

The reviewed tariff has no explicit Markdown or Math Markdown row. Printed-text or handwriting prices are not silently assigned.

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

Why it is included

Provider names this native recognition or synthesis model in its hosted API documentation. No exact matching downloadable checkpoint was found in the reviewed official Yandex model inventory; public SDK source does not release the served model weights.

Lifecycle dates

Sources reviewed