Pricing
Overview
Perplexity AI is introducing new pricing with search modes, which will be available starting today. However, the current (legacy) pricing remains active and will be deprecated on 04/18/2025.
Key Changes
- New search modes (
high,medium,low) allow for greater control over search costs. - By default, models will continue using legacy pricing unless you explicitly specify a search mode in your API call.
- Legacy pricing will be deprecated on 04/18/2025. Please update your integration accordingly.
Legacy Pricing (will be deprecated on 04/18/2025)
| Model | Input Tokens (Per Million) | Reasoning Tokens (Per Million) | Output Tokens (Per Million) | Price per 1,000 Requests |
|---|---|---|---|---|
sonar-deep-research | $2 | $3 | $8 | $5 |
sonar-reasoning-pro | $2 | - | $8 | $5 |
sonar-reasoning | $1 | - | $5 | $5 |
sonar-pro | $3 | - | $15 | $5 |
sonar | $1 | - | $1 | $5 |
r1-1776 | $2 | - | $8 | - |
📌 Important Note: If you do not specify a search mode in your API call, billing will default to this legacy pricing.
Detailed Pricing Breakdown (for “Legacy” pricing)
New Pricing (Search Modes)
The new pricing structure introduces search context modes:
- High: Uses the most search context for richer results.
- Medium: Balanced search for cost vs. depth.
- Low: Minimal search context, optimized for cost savings.
Non-Reasoning Models
Models optimized for fast, cost-effective search and information retrieval.
Sonar Pro
Advanced search model optimized for complex queries and deeper content understanding.
Learn more →
Pricing:
| Metric | High | Medium | Low |
|---|---|---|---|
| Input Tokens (Per Million) | $3 | $3 | $3 |
| Output Tokens (Per Million) | $15 | $15 | $15 |
| Price per 1000 Requests | $14 | $10 | $6 |
Sonar
Lightweight, cost-effective search model designed for quick, grounded answers.
Learn more →
Pricing:
| Metric | High | Medium | Low |
|---|---|---|---|
| Input Tokens (Per Million) | $1 | $1 | $1 |
| Output Tokens (Per Million) | $1 | $1 | $1 |
| Price per 1000 Requests | $12 | $8 | $5 |
Reasoning Models
Models optimized for multi-step reasoning, problem-solving, and real-time search.
Sonar Reasoning Pro
Enhanced reasoning model with multi-step problem-solving capabilities and real-time search.
Learn more →
Pricing:
| Metric | High | Medium | Low |
|---|---|---|---|
| Input Tokens (Per Million) | $2 | $2 | $2 |
| Output Tokens (Per Million) | $8 | $8 | $8 |
| Price per 1000 Requests | $14 | $10 | $6 |
Sonar Reasoning
Quick problem-solving and reasoning model, ideal for evaluating complex queries.
Learn more →
Pricing:
| Metric | High | Medium | Low |
|---|---|---|---|
| Input Tokens (Per Million) | $1 | $1 | $1 |
| Output Tokens (Per Million) | $5 | $5 | $5 |
| Price per 1000 Requests | $12 | $8 | $5 |
Deep Research Models
Models designed for exhaustive research, expert-level analysis, and detailed report generation.
Sonar Deep Research
Best suited for exhaustive research, generating detailed reports and in-depth insights.
Learn more →
Pricing:
| Metric | Cost |
|---|---|
| Input Tokens (Per Million) | $2 |
| Output Tokens (Per Million) | $8 |
| Price per 1000 Requests | $5 |
| Reasoning Tokens (Per Million) | $3 |
Offline Model
For private, factual-based answering without real-time web search.
r1-1776
Offline chat model that does not use search but provides local AI capabilities.
Learn more →
Pricing:
| Metric | Cost |
|---|---|
| Input Tokens (Per Million) | $2 |
| Output Tokens (Per Million) | $8 |