Schematron V2 Turbo is Inference.net’s high-throughput HTML-to-JSON extraction model, designed to turn messy HTML into schema-conforming JSON using schema-constrained decoding. It is part of the Schematron family and is described as optimized for speed and cost at scale.
The model is presented as a 3B-parameter extraction model for web scraping and data ingestion workflows, with structured output support and strict JSON generation by construction.
| Developer | Inference.net |
|---|---|
| Model ID | inference-net/schematron-v2-turbo |
| Release Date | September 12, 2026 |
| Context Window | 128,000 tokens (~128K) |
| Max Output Tokens | 8,192 tokens |
| Modalities | Text → Text |
| Capabilities | streaming, structured (JSON schema) output |
Schema-first extraction
You supply a JSON Schema or typed schema such as Pydantic or Zod, and the model returns JSON that conforms to that schema.
Strict JSON output
Inference.net describes the model as using strict JSON mode and schema-constrained decoding so output is valid JSON by construction.
Optimized for throughput
The Turbo variant is positioned as the throughput- and cost-optimized sibling in the Schematron family.
Long-context HTML handling
The model is documented with a 128K-token context window for processing long or noisy HTML inputs.
Structured outputs support
Catalog listings identify structured output support for the model.
Web scraping
Suitable for extracting structured records from messy web pages into schema-conforming JSON.
Data ingestion
Designed for pipelines that convert arbitrary HTML documents into typed data for downstream systems.
High-volume extraction
Recommended where speed, latency, and cost matter more than maximum extraction quality.
Schematron V2 Small
Inference.net describes Schematron V2 Small as the better choice for complex schemas or very long pages, while Turbo is tuned for maximum throughput and lowest cost at scale.
Schematron 3B
Inference.net states that requests sent to the older Schematron 3B route are now automatically routed to Schematron V2 Turbo.
Sources: https://inference.net/blog/schematron-v2/ · https://inference.net/models/schematron-v2-turbo/ · https://developer.puter.com/ai/inference-net/schematron-v2-turbo/
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
High-throughput HTML-to-JSON extraction is a documented use case for Schematron V2 Turbo, especially when cost and latency matter.
It is suited to web scraping pipelines that need clean, typed JSON from messy HTML with schema adherence.
It is also documented for product data ingestion and converting arbitrary web pages into structured data.
For latency-sensitive crawling and scraping workflows, the model is described as a fit for financial document parsing for live trading and real-time price monitoring.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "inference-net/schematron-v2-turbo", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "inference-net/schematron-v2-turbo", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"inference-net/schematron-v2-turbo","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
It is a high-throughput HTML-to-JSON extraction model for turning messy HTML into schema-conforming JSON. It is designed for web scraping and data extraction pipelines that need strict structured output.
Yes. The model supports structured outputs and schema-constrained decoding, so output is valid JSON by construction.
Yes. The model supports streaming.
Send raw or cleaned HTML together with a JSON Schema, or a typed schema definition such as Pydantic or Zod. The model is built to extract data that conforms to the schema you provide.
The model uses strict JSON mode and schema-constrained decoding. The vendor states that it produces 100% schema-compliant output.
Yes. It is specifically described as being purpose-built for messy HTML and robust to noisy markup.
It produces clean, typed JSON rather than free-form narration. The model is optimized for schema-first extraction workflows.
It fits web scraping, product data ingestion, and document-to-JSON extraction workflows. The emphasis is on speed and cost at scale.
Schematron V2 Turbo is the throughput-optimized sibling, intended for faster processing and lower cost at scale. The documentation also says it exceeds the quality of the previous-generation Schematron 3B while staying focused on high-throughput extraction.
No. The vendor states the endpoint does not support tool calling.
