Schematron V2 Turbo

LLMBase's third-party catalog profile explicitly names Inference.net's Schematron V2 Turbo under the exact slug "inference-net/schematron-v2-turbo," confirming it as a 3B-parameter text-only model purpose-built for HTML-to-JSON extraction that requires extraction instructions to be supplied via a JSON schema in the res The page notes that Schematron V2 Turbo is not currently available in LLMBase's own chat or inference API, positioning this entry as a research/comparison data point rather than a primary Inference.net announcement. As an aggregator restatement of baseline specifications rather than breaking news, the profile is useful

Infron's performance/pricing page explicitly names Inference.net's Schematron V2 Turbo under the slug "inference-net/schematron-v2-turbo" and lists it as a text-generation, streaming, tool-calling, JSON-mode capable model with a single official service tier priced at $0.03 per million input tokens, $0.15 per million ou The page presents uptime (100.00% over 72 hours after routing/fallbacks) and provider details for Inference.net alongside links to explore a sibling model, Schematron V2 Small, which is also attributed to Inference.net. This is third-party routing/serving-layer data rather than a primary Inference.net announcement, and

A timeline of AI model releases lists "Inference.net: Schematron V2 Turbo" as the top entry, noting an API launch on 2026-09-12 with a 128k context window. The page is auto-refreshed from the OpenRouter models API and presents a chronological feed of new model additions, providing a concrete launch date and context len No benchmarks, pricing, architectural details, or capability claims for Schematron V2 Turbo are provided on the page, and no first-party Inference.net or OpenRouter announcement is linked. A sibling entry for "Inference.net: Schematron V2 Small" (also 128k context, same date) appears as entry #02, situating the Turbo v

Inference.net announced Schematron V2 on April 16, 2026, framing it as a frontier HTML-to-JSON extraction model positioned at a fraction of the cost of larger alternatives. The blog post marks the second generation of Inference.net's schema-driven extraction family, emphasizing turnkey infrastructure for production deployments. The V2 release targets developers who need reliable structured extraction from raw HTML, supporting JSON-schema response formats and long-context inputs suited for large documents and transcripts. While the post covers the V2 family broadly, the Turbo variant is positioned within this lineup for throughput-prioritized extraction workloads on Inference.net's managed inference platform.

Inference.net released Schematron V2 Turbo, a 3B-parameter model specialized for HTML-to-JSON extraction, as cataloged by the Benchable benchmark index dated September 2026. It exposes a 128K-token context window and requires extraction instructions supplied through a JSON schema in response format rather than natural-language prompting. The model targets high-throughput extraction workloads, priced at $0.03 per 1M input tokens and $0.15 per 1M output tokens, with structured outputs and response-format parameters supported. Benchable benchmarks report a 99% reliability success rate and top-tier speed, but very low general-task scores including 0% instruction following and 14% reasoning accuracy, consistent with a narrow specialist design.