# Interfaze Beta

Interfaze Beta is a hybrid AI system from Interfaze designed for developer tasks that require consistent, structured output. It routes each request to a specialized DNN or CNN model where one fits, and falls back to an LLM otherwise. The result is one OpenAI-compatible endpoint that covers OCR, web extraction, classification, speech-to-text, translation, structured output, and text-to-SQL.

- **Model ID:** `interfaze/interfaze-beta`
- **Type:** chat
- **Providers:** interfaze
- **Context window:** 1,000,000
- **Maximum output tokens:** 32,000
- **Pricing:** $1.5/1M input tokens, $3.5/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/interfaze-beta

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

```ts
import { streamText } from 'ai'

const result = streamText({
  model: 'interfaze/interfaze-beta',
  prompt: 'Why is the sky blue?'
})
```

## About

Interfaze built Interfaze Beta around a routing layer. Each request goes to whichever specialized model fits the task. Small CNN and DNN models handle perception work like OCR and object detection. An LLM handles language reasoning. Custom tools cover web search, a code sandbox, and configurable safety guardrails. The endpoint is a single OpenAI-compatible URL.

The context window is 1M tokens and maximum output is 32K tokens. Interfaze reports 70.7% on OCRBench V2 for the native OCR path and 98 to 99% accuracy on structured output generation. Inputs accepted include text, images, audio, files, and video. Reasoning is available for harder queries.

Task coverage includes OCR and document extraction, object detection driven by natural language prompts, web scraping (with handling for sites that block bots), speech-to-text with speaker diarization, translation across many languages, classification, structured output, text-to-SQL, and multimodal question answering.

Because the underlying mix of CNNs, DNNs, and an LLM stays opaque behind one endpoint, integration looks identical to any other chat-completions model. Send a prompt with optional image, audio, or file attachments and get back a response that matches your requested schema. See  for product documentation and https://interfaze.ai/ for the model page.

## What to consider

Interfaze Beta is a beta release. It targets workloads where deterministic, parseable output matters more than open-ended conversation. If your pipeline lives or dies by JSON schema adherence, OCR fidelity, or reliable web scraping, Interfaze Beta is built for that shape. For free-form chat, code generation, or general reasoning, a general-purpose frontier model usually fits better.

Call Interfaze Beta through the AI SDK, Chat Completions API, Responses API, Messages API, or other API formats, from TypeScript or Python. Inputs cover text, images, audio, files, and video.

## When to use

### Best For

- Document pipelines that need OCR and field extraction with high accuracy
- Structured output generation where downstream systems parse the response against a schema
- Web scraping and extraction workflows, including sites that block typical bots
- Multimodal classification tasks across images, audio, and documents
- Text-to-SQL and natural-language data querying

### Consider Alternatives When

- You need a general-purpose chat or code generation model
- Open-ended reasoning, long-form writing, or creative generation is the main workload
- You require a stable production release rather than a beta model

## Best for

- **OCR Pipelines:** Document extraction workloads that need high field-level accuracy on scanned content
- **Structured Output:** API responses that downstream systems parse against a strict schema
- **Web Extraction:** Scraping workflows, including sites that block typical bots
- **Multimodal Classification:** Tasks that mix images, audio, files, or video alongside text
- **Text-to-SQL:** Natural-language queries translated into runnable SQL

## Consider alternatives

- **General Chat:** A general-purpose frontier model fits open-ended conversation better
- **Code Generation:** A coding-tuned model handles long code synthesis with more depth
- **Production Stability:** A stable release is a safer pick than a beta for critical paths
- **Creative Writing:** A model tuned for long-form prose serves narrative work more naturally

## Frequently asked questions

### What is Interfaze Beta?

Interfaze Beta is a hybrid AI system from Interfaze that routes each request to a specialized DNN or CNN model when one fits, and falls back to an LLM otherwise. It targets developer tasks like OCR, scraping, classification, structured outputs, and web extraction.

### What is the context window and output limit?

The context window is 1M tokens and the maximum output is 32K tokens.

### Which input modalities does Interfaze Beta support?

Text, images, audio, files, and video. The API stays OpenAI Chat Completions compatible across all of them.

### How do I call Interfaze Beta through AI Gateway?

Set the model to `interfaze/interfaze-beta` in the AI SDK, Chat Completions API, Responses API, Messages API, or other API formats, from TypeScript or Python. AI Gateway handles authentication and routing. See https://interfaze.ai/ for the model page.

### What is the pricing?

On AI Gateway, Interfaze Beta costs $1.5 per million input tokens and $3.5 per million output tokens. Current rates appear on this page.

### How well does Interfaze Beta handle structured output?

Interfaze reports 98 to 99% accuracy on structured output generation, which makes Interfaze Beta a fit for pipelines that parse responses against a schema.

### Does Interfaze Beta support zero data retention?

Zero Data Retention is not currently available for this model. Zero Data Retention is offered on a per-provider basis. See https://vercel.com/docs/ai-gateway/capabilities/zdr for details.

## Links

- [Model page](https://vercel.com/ai-gateway/models/interfaze-beta)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://interfaze.ai/)