Relace Search
Neural Network
Relace Search is a Relace model for agentic code search: it finds files relevant to a query via parallel tool calls.
Max answer length
(in tokens)
Context size
(in tokens)
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Relace Search?
Relace Search is a model by Relace for agentic codebase search. It does not build a vector index like RAG; instead, it runs 4 to 12 parallel view_file and grep calls, explores the repository itself, and returns files relevant to the query. This approach is closer to how a developer searches: the model relies on file content rather than embedding proximity. In BotHub, it is available without a VPN or foreign card — you pay with a Russian card in rubles on a pay-as-you-go basis, without a subscription. Nearby, in the same window, there are 250+ other neural networks, and a unified OpenAI-compatible API allows you to connect Relace Search to your agent and switch models if desired without rewriting the integration; traffic is encrypted, and companies have access to contracts, invoices, EDI, and an admin panel with limits. It is useful for quickly finding where a feature is implemented before editing; gathering context for a code agent that then writes a patch; understanding someone else's legacy repository; understanding which files a refactoring will affect; and linking code search with a task tracker to resolve tickets faster.Frequently asked questions about Relace Search
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.
The Relace model works with text and documents, accepts images as input, and can call tools. It is suitable for searching large codebases and files, analyzing documentation, navigating projects, and gathering context for further work with code.
The output limit is up to 128,000 tokens per response, which is enough for detailed summaries of multiple files, long code snippets, or detailed documentation analysis. The context is 256,000 tokens, so a large volume of source code will fit into the request.
In our data, a separate reasoning mode for this model is not noted. However, you can ask it to break down a task into steps directly in the prompt, and if you need explicit chains of thought, it is easy to switch to a reasoning model in BotHub in the same window.
Yes, the model supports function calling and structured JSON output. This allows you to embed it into agentic scenarios, pass file search results further into your code, and receive a predictable format that is convenient for parsing.
Text, documents, and images are accepted as input, so interface screenshots, architecture diagrams, and PDF documentation can be analyzed. Audio and video are not noted in our data; for those, there are separate models in BotHub in the same interface.
We do not have specific data on the quality of Russian for this model, so we will not make any promises. Test it on your task: in BotHub, payment is pay-as-you-go for tokens, without a subscription, and if necessary, you can compare the result with another model.