Inferse
For vendors
AdvertiseTop spots, featured rows, home sponsor Get listedAdd a product we don't cover yet Claim a listingFree: verify the domain, then correct your page Vendor dashboardNumbers, orders and placements
List your platform
Vendor dashboard

Sonar Deep Research

Perplexity · released 7 Mar 2025

Input$2per 1M tokens
Output$8per 1M tokens
Context128Ktokens
WeightsClosed
context on a 1K–2M scale

About

Sonar Deep Research is Perplexity's research-focused model for multi-step retrieval, synthesis and reasoning across complex topics. It autonomously searches for, reads and evaluates sources, then refines its approach as it gathers information. The model accepts text and produces text, with reasoning support listed. Its context window is 128,000 tokens, and its maximum output is 115,200 tokens. Input costs $2 per 1M tokens and output costs $8 per 1M tokens. It was released on 2025-03-07. The listed official page is on OpenRouter. These facts make the model relevant to research workflows involving source gathering and synthesis, while its context, output ceiling and per-token rates are useful for planning longer tasks.

Who it is for

It suits teams building research workflows that need multi-step retrieval, source evaluation and synthesis. Its listed interface is text in and text out.

What is good

  • Designed for multi-step retrieval and synthesis.
  • Autonomously searches, reads and evaluates sources.
  • 128,000-token context window.
  • Maximum output is 115,200 tokens.

What to know first

  • Input costs $2 per 1M tokens.
  • Output costs $8 per 1M tokens.
  • Inputs and outputs are text only.

Verdict

Sonar Deep Research is oriented toward gathering and synthesizing information across complex topics. Its token prices and text-only interface are important considerations for research workflows.

Details

Lab
Perplexityopenrouter.ai · 3 Oct 2026
Context
128Kopenrouter.ai · 3 Oct 2026
Input price
$2 / 1Mopenrouter.ai · 3 Oct 2026
Output price
$8 / 1Mopenrouter.ai · 3 Oct 2026
Max output
115,200 tokensopenrouter.ai · 3 Oct 2026
Inputs
textopenrouter.ai · 3 Oct 2026
Open weights
Noopenrouter.ai · 3 Oct 2026

More longest-context models

See the list
Auto Router 2,000K Context Auto Router (Beta) 2,000K Context Grok 4.20 2,000K Context Grok 4.20 Multi-Agent 2,000K Context Pareto Code Router 2,000K Context Llama 4 Scout 1,311K Context Open source

Listed on Inferse

Sources