Schematron V2 Turbo
Neural Network
Schematron V2 Turbo is an Inference.net 3B parameter model for extracting data from HTML to JSON via API.
Specifications
- Input
- text, documents
- Output
- text
- Context size
- 128 000 tokens
- Max answer length
- 8 192 tokens
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Schematron V2 Turbo?
Schematron V2 Turbo is a compact 3B parameter model by Inference.net, designed for one task: converting raw HTML into structured JSON. The Turbo version focuses on throughput, making it suitable for streaming large volumes of pages. Extraction instructions are set via a JSON schema in the response_format parameter rather than prompt text, ensuring predictable responses that require minimal post-processing. In BotHub, the model works without VPN or foreign cards: pay with Russian cards in rubles for actual tokens used, with no balance expiration. Connection is via a unified OpenAI-compatible API, allowing you to switch to any of over 250 models in the same code without rewriting integration. For companies, we offer contracts, invoices, EDI, and an admin panel with limits. It is useful for scraping product cards and competitor prices, building ETL pipelines, populating databases from web sources, converting page archives into datasets, extracting contacts from corporate sites, or enriching CRM data from open catalogs.Frequently asked questions about Schematron V2 Turbo
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.
This is a text model with a 128,000 token context that accepts both text and documents. It is well-suited for analyzing large files, data extraction, summarization, report preparation, and working with long instructions. Via BotHub, you can compare it with other models in one window.
According to our data, the maximum response volume is 8192 tokens. This is enough for detailed reports, document analysis, and long drafts. If you need more text, it is logical to split it into several requests or refine it in parts.
A separate reasoning mode before answering is not noted in our data, so we cannot confirm it. The model builds the response to your request as a whole. If step-by-step logic is needed, help it with the phrasing: ask it to break down the task into steps and justify the conclusions.
Support for function calling and structured JSON output is not noted in our data, so we cannot guarantee it. The model works with text and documents. If your scenario requires a strict format, define the structure directly in the prompt and check the result with test requests.
According to our data, the model accepts text requests and documents. We have no information about uploading images, audio, or video, so we will not make any claims. If you need to process an image or media file, check out the corresponding model types in BotHub.
We have no data on the quality of Russian language support for this model, so we cannot promise anything specific. It is known that it works with text and documents. The easiest way to test it on Russian-language tasks is with short test requests: payment in BotHub is per token used, no subscription required.