Switchyard
Neural Network
Switchyard is an open-source model router that automatically distributes requests between neural networks and optimizes their cost.
Specifications
- Input
- text, documents
- Output
- text
- Context size
- 1 000 000 tokens
- Max answer length
- 330 000 tokens
import OpenAI from 'openai';
const openai = new OpenAI({
apiKey: '<your bothub access token>',
baseURL: 'https://openai.bothub.chat/v1'
});
// Sync - Text generation
async function main() {
const chatCompletion = await openai.chat.completions.create({
messages: [{ role: 'user', content: 'Say this is a test' }],
model: 'switchyard',
});
}
// Async - Text generation
async function main() {
const stream = await openai.chat.completions.create({
messages: [{ role: 'user', content: 'Say this is a test' }],
model: 'switchyard',
stream: true
});
for await (const chunk of stream) {
const part: string | null = chunk.choices[0].delta?.content ?? null;
}
}
main();
How it works Switchyard?
Switchyard is an open-source request router: instead of a single fixed neural network, it selects the model for each specific request based on market demand data, keeping costs down. In BotHub, the model works from Russia without a VPN or foreign card; payment is made with Russian cards per token, without a subscription, and unused balances do not expire. With over 250 neural networks available in one window, it's easy to compare the router's response with a manually selected model, and the unified OpenAI-compatible API allows you to switch models in your integration without rewriting code. Chats are encrypted via AES-GCM, data is not saved, and companies have access to contracts, invoices, electronic document management (EDO), and an admin panel with limits. It is useful if you need to reduce the cost of repetitive requests, build a prototype of a multi-model service, handle first-line support, perform draft generation and text editing, or test which neural network handles your task most accurately.Frequently asked questions about Switchyard
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.
Switchyard is a text model from NVIDIA. It accepts text and documents as input, making it suitable for working with long materials: analyzing reports, summarizing and structuring, editing and rewriting, answering questions about file content, and assisting with code and technical documentation.
The maximum volume of a single response is 330,000 tokens, which is hundreds of pages. This capacity is enough for a long report, translating a large document, or detailed technical documentation in its entirety, without manual splitting. The context window is one million tokens.
A separate reasoning mode is not noted in our data, but that doesn't mean it doesn't exist. If you need step-by-step explanations, ask for them directly in the request: the model will break down the task step-by-step within the response itself — this is usually sufficient for logic analysis or calculations.
There are no notes about function calling or strict structured output in our data, so we cannot guarantee support. In practice, you can specify the format directly in the prompt and ask it to return only JSON. Before production, test the model on your scenarios via the unified BotHub API.
Documents, yes, this is a confirmed capability: upload files and ask questions about them. Images, audio, and video as input formats are not noted in our data. If you need to work with media, multimodal models are available nearby in BotHub, and switching happens in one window.
We do not have specific data on the quality of Russian for this particular model, so we will not make any promises. The easiest way is to test it with your own text: open Switchyard on the website or in the BotHub Telegram bot. Access from Russia without a VPN or foreign card, payment in rubles.