Jev Router
Neural Network
Jev Router by TypeSafe is a request router that automatically selects the model and reasoning depth for tasks, available in BotHub.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 1 000 000 tokens
- Max answer length
- 65 536 tokens
- Reasoning effort
- low, medium, high
- Capabilities
- reasoning, function calling
How it works Jev Router?
Jev Router is a routing layer by TypeSafe: it analyzes requests and decides which model and reasoning effort level to use to balance quality, speed, and cost. Based on Jev, TypeSafe's first System One model, the routing adapts to your request type: simple tasks don't waste resources, while complex ones get deeper analysis. Via BotHub, the router works from Russia without VPN or foreign cards; pay with Russian cards in rubles only for tokens actually used—no subscription needed, balance doesn't expire. Over 250 models are available in the same window for easy comparison, and a unified OpenAI-compatible API allows replacing models in your product without rewriting integrations. Requests are encrypted via AES-GCM, and companies get contracts, invoices, EDI, and an admin panel with limits. Useful for processing diverse requests in support and chatbots, writing and refactoring code with debugging, analyzing documents and analytics, drafting texts, and in multi-model services where guessing the right model manually is inefficient.Frequently asked questions about Jev Router
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.
It is a text model with a million-token context: it analyzes large documents and entire codebases, writes and refactors code, and helps with debugging, analytics, and STEM tasks. Suitable for research, working with voluminous documentation, and building agent scenarios via API.
The output limit is up to 330,000 tokens per response, enough for voluminous documents, detailed technical breakdowns, long code listings, or translating large texts entirely without manually splitting tasks into dozens of small requests.
Yes, the model has a reasoning mode: before providing a result, it builds a sequence of steps. This significantly helps with math, logic, code analysis, and multi-step tasks where it is important to derive the answer carefully rather than guessing.
Yes, function calling and structured JSON output are supported. The model can be connected to external tools and databases, used to build agents, and return predictable answers based on a schema. In BotHub, this is done via a unified OpenAI-compatible API.
In addition to text, images and documents are accepted as input: you can send an interface screenshot, diagram, chart, PDF, or contract and ask to analyze the content, extract data, and draw conclusions. The model generates text.
The model works with text in a general way, so it is most practical to test it on your specific task. In BotHub, this is convenient: all models are available in one window, and you can compare multiple responses, paying only for the tokens actually used.