Schematron V2 Small
Neural Network
Schematron V2 Small is an Inference.net 3B parameter model for extracting data from HTML into structured JSON based on a schema.
Specifications
- Input
- text, documents
- Output
- text
- Context size
- 128 000 tokens
- Max answer length
- 4 096 tokens
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Schematron V2 Small?
Schematron V2 Small is a compact 3 billion parameter model by Inference.net that converts HTML pages into structured JSON. Instructions are defined by a JSON schema, ensuring predictable output rather than free text. It focuses on extraction quality for complex schemas and long pages. In BotHub, it is available without a VPN or foreign card, with payment via Russian cards in rubles and pay-as-you-go. Requests go through a unified OpenAI-compatible API, so you can connect this model or switch to any of the 250+ others without rewriting your integration. For companies, we offer contracts, invoices, and an admin panel with limits. It is useful for collecting product catalogs and prices from websites, consolidating job or ad listings into a single table, migrating content from old CMS, normalizing parser output before database entry, or preparing structured data for RAG and analytics.Frequently asked questions about Schematron V2 Small
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
schematron-v2-small is a text model from inference-net that also works with documents. The 128,000 token context window allows loading long files entirely, making the model suitable for data extraction, structuring, summarization, and parsing of large texts and technical documentation.
The single response limit is 4096 tokens, which is several pages of text. This is sufficient for summaries, structured exports, and document-based answers. If you need more volume, break the task into parts or ask it to continue: the 128,000 token input allows it to keep the entire material in view.
In our data, a separate reasoning mode for schematron-v2-small is not noted, so we do not promise step-by-step reasoning. In practice, you can ask it to break down the task step-by-step directly in the prompt. If reasoning is critical, switch to another model in BotHub — you won't need to rewrite the integration.
Support for function calling and strict JSON mode is not noted in our data. You can describe the required response structure directly in the prompt and ask it to return only JSON, then verify the result with a validator. If a schema is mandatory, use a suitable model via BotHub's unified OpenAI-compatible API.
Besides text, schematron-v2-small accepts documents: upload a file and ask questions about it or create a structured summary. Input for images, audio, and video is not noted in our data — multimodal models are available for them in the same BotHub window.
We have no confirmed data on the query and response languages of schematron-v2-small, so it is safer to test it with your own material: send a couple of typical documents and compare the result with another model. Access from Russia is available without a VPN or foreign card, with pay-as-you-go billing for tokens.