← All models

Sonar Reasoning Pro

Language Search-grounded Deprecated Full comparison ↗

perplexity/sonar-reasoning-pro · by Perplexity · decoder-only-transformer

Search-grounded model. Unlike a standard base model, this offering combines an underlying LLM with live web retrieval and citations — not a directly comparable like-for-like with a non-search-grounded entry. Billed with the token pricing below plus a per-request search-context fee (see Pricing).

Pricing — 1 offering(s)

Perplexity

About this offering: Perplexity ended support for Sonar models on Chat Completions on 2026-09-27, but it still serves those requests while it moves each model to its Agent API. Its pricing page no longer shows the Sonar tables, so the prices here are the last published rates.

Input tokens

  • $2.00 / 1M tokens (input) Current 2026-07-26 → present

Output tokens

  • $8.00 / 1M tokens (output) Current 2026-07-26 → present

Search context fee (billed on top of token pricing, per query)

  • low $6.00 / 1k requests
  • medium $10.00 / 1k requests
  • high $14.00 / 1k requests

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

See how Sonar Reasoning Pro fits into a cost-aware routing setup

See how →

Capability profile

How Sonar Reasoning Pro rates across core capability dimensions, with the task-level evidence behind each rating.

reasoning Strong

Perplexity positions this as its "precise reasoning offering with Chain of Thought", best for "complex analyses requiring step-by-step thinking, strict instruction adherence, and logical problem-solving" (docs models page). The strongest reasoning in the Sonar family short of Deep Research.

tool use Moderate

Native real-time web search with citations, integrated into the reasoning loop. The registry entry sets `agentic_tool_use: false` — it is not a general agentic/function-calling model, the search is built in.

instruction following Strong

Perplexity calls out 'strict instruction adherence' as a strength of this variant specifically.

context window Moderate

128,000 tokens.

speed Weak

Chain-of-thought generation plus search makes it the slower, more expensive path; Perplexity explicitly says it is "not ideal for ... speed-critical applications."

cost efficiency Moderate

$2 / 1M input, $8 / 1M output (cheaper output than Sonar Pro), plus the $6-$14 per 1,000 requests search-context fee. The CoT tokens are billed as ordinary output tokens on this variant (unlike Deep Research, which has a separate reasoning-token category).

multilingual Moderate
coding Moderate

Reasoning base helps with code analysis / debugging-style questions; not marketed as a coding model.

Ratings are estimated — limited independent data is available for this model.

Operator guidance

Route here when the task genuinely needs chain-of-thought analysis on top of cited web search and can tolerate higher latency. For simple factual lookups it is slower and dearer than Sonar with no benefit. For exhaustive hundreds-of-sources reports, Sonar Deep Research is purpose-built and prices its citation/reasoning tokens separately. Outside Perplexity, pairing a standalone reasoning model (o-series, DeepSeek R1, Claude with extended thinking) with your own retrieval gives more control but loses the tuned grounding.

Use cases

Limitations

Citations