Sonar Deep Research
Neural Network
Sonar Deep Research is a Perplexity model for deep research: multi-step search, source evaluation, and synthesis of findings.
Max answer length
(in tokens)
Context size
(in tokens)
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Sonar Deep Research?
Sonar Deep Research is a Perplexity research model that works through a series of queries rather than a single one: it independently searches for materials, reads and evaluates findings, refines its direction during data collection, and synthesizes everything into a coherent answer on complex topics. Instead of relying solely on memory, it uses multi-step information retrieval and synthesis, making it ideal for broad topic analysis and accurate reasoning based on gathered sources. In BotHub, this model is accessible from Russia without a VPN or foreign card: pay in rubles for tokens used, and unused balances do not expire. Over 250 other neural networks are available in the same window, allowing you to switch between them and compare answers for the same task. There is a unified OpenAI-compatible API for products, and chats are encrypted via AES-GCM. Analysts can use it for market and competitor overviews, marketers to understand a niche before launch, product teams to consolidate scattered materials into a report, researchers to prepare literature reviews, and consultants to quickly get up to speed in an unfamiliar industry.Frequently asked questions about Sonar Deep Research
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.
Perplexity's sonar-deep-research is designed for deep research: the model searches the web, reasons about findings, and compiles a detailed analytical answer. It is useful for market overviews, analyzing complex topics, and working with documents and images—both are accepted as input.
The output limit is 115,200 tokens per response, which is enough for a very long analytical report. Plus, there is a 200,000-token context window: you can upload a large array of materials and get a detailed analysis without breaking the task into parts.
Yes, reasoning before answering is confirmed. The model builds a chain of thought, verifies data found online, and only then formulates a conclusion. Therefore, the answer is not instantaneous, but for complex research queries, it is noticeably more reasoned and supported by sources.
We have no data indicating support for function calling or structured JSON, so we cannot guarantee it. You can usually request a JSON format directly in the prompt, but for strict schemas, BotHub offers other models—the API is unified and OpenAI-compatible, so you won't need to rewrite your integration.
Images and documents—yes, they can be uploaded along with text. Audio and video are not marked in our data, so for audio or recording analysis, it is more convenient to switch to a specialized model: all are available in BotHub in one window.
We do not have specific data on this model's Russian language capabilities, so we cannot make any promises—it is easier to test it with your own task. Access to sonar-deep-research from Russia works without a VPN or foreign card, with payment via Russian cards in rubles, based on actual token usage.