Sonar Reasoning Pro
Neural Network
Sonar Reasoning Pro is a Perplexity reasoning model with chain-of-thought for analytical tasks supported by web search.
Max answer length
(in tokens)
Context size
(in tokens)
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Sonar Reasoning Pro?
Sonar Reasoning Pro is a Perplexity reasoning model: it breaks down tasks step-by-step via chain-of-thought and relies on Perplexity's built-in search. This approach is essential when analysis matters more than an instant answer: verifying sources, checking hypotheses, and multi-step reasoning. Via BotHub, the model works without a VPN or foreign card: you pay with a Russian card in rubles only for tokens used, with no subscription, and unused tokens do not expire. With over 250 neural networks in one window, it's easy to compare answers, and the unified OpenAI-compatible API allows switching models without rewriting integrations; requests are encrypted via AES-GCM, and data is not stored. Analysts can use it to gather and verify facts, marketers to analyze markets and competitors using sources, developers to debug logic step-by-step, and researchers to review publications. For corporate clients, BotHub provides contracts, invoices, EDI, and an admin panel with limits.Frequently asked questions about Sonar Reasoning Pro
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.
sonar-reasoning-pro by Perplexity is a model with reasoning and native web search. It is suitable for analytics, research, document analysis, and questions requiring fresh web data. It accepts text, images, and documents as input, and the 128,000-token context allows working with large materials.
Up to 115,200 tokens in a single response — enough for a long analytical review, detailed document analysis, or comprehensive report. The total window is 128,000 tokens, so the more input material you provide, the less space remains for output.
Yes, reasoning before answering is confirmed: the model builds a chain of thought before formulating a conclusion. Combined with native web search, this helps in tasks where it is important to compare multiple sources and show the logic of the solution, rather than just providing a short reply.
In our data, function calling and structured JSON output are not noted, so we cannot guarantee them. In practice, you can request the desired structure directly in the prompt and check the result. If strict tool calling is required, you can easily switch to another model via the unified BotHub API.
The model accepts images and documents as input: send a screenshot, diagram, or file and ask to analyze the content. Audio and video input are not supported. As for output, do not expect text, image, audio, or video generation.
There are no language-specific characteristics in our data, so we cannot promise anything specific — it is easier to test the model on your task directly in BotHub. Access works from Russia without a VPN or foreign card, with payment via Russian cards in rubles based on actual usage.