Sonar (Agent API) API Pricing
Perplexity ended its Sonar Chat Completions API on September 27, 2026, and perplexity/sonar on the Agent API is now the one model it sells per token under its own name: $0.25/M input, $2.50/M output and $0.0625/M cache-hit input. That is a different bill from the old Sonar row's $1/$1 plus a per-request search fee: on the Agent API, searches and other tools are metered per call, separately from tokens. Checked by hand against docs.perplexity.ai.
Run the numbers.
Live calculator pre-loaded with the Agent API's token rates for perplexity/sonar. It prices tokens only: every tool the run calls is billed per invocation on top - web_search is $0.0025 a call ($0.001 with Fast Search) and fetch_url $0.0005 - so add those for search-heavy work.
Real-world presets.
Web answer, tokens only
Market check, tokens only
Evidence-rich answer
Daily topic monitor
This is a chars-per-token approximation, not a real tokenizer. Actual tokens vary by language, code density, and tool-call overhead — counts are typically ±10–20% off for English prose, more for code or non-Latin scripts. For exact billing, use the vendor's official tokenizer.
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| Sonar (Agent API) Current | $0.25 cache $0.0625 | $2.50 | $0.289 agentic 92/8 | not listed | Perplexity's own model, billed per token |
| Sonar | $1.00 | $1.00 | $1.00 retired | not listed | The Sonar API row, retired Sep 27 |
| Sonar Pro | $3.00 | $15.00 | $3.96 retired | not listed | Retired Sonar Pro, for invoice checks |
| GPT-6 Luna | $0.10 cache $0.01 | $0.50 | $0.0682 −4.2× | 1.05M | Low-cost OpenAI model on the Agent API |
| Gemini 3.8 Flash | $0.75 cache $0.075 | $3.75 | $0.481 +67% · promo | 1M | Mainstream multimodal work |
| Reka Flash | $0.80 | $2.00 | $0.896 +3.1× | not listed | Fast chat and production assistants |
Frequently asked.
Practical questions about Perplexity's per-token model after the Sonar API ended - what it costs, what it does not include, and how it relates to the old Sonar rows.
Q · 01 What does Sonar cost on the Agent API? +
perplexity/sonar at $0.25/M input, $2.50/M output and $0.0625/M cache read. There is no per-request fee on the model itself; tools the run calls are billed per invocation.Q · 02 How much do searches add? +
web_search is $0.0025 per invocation with standard search and $0.001 with Fast Search, fetch_url $0.0005, people_search and finance_search $0.005, and a code sandbox $0.03 per session. On a short answer that is more than the tokens: 1,000 input and 500 output tokens cost about $0.0015 here, one standard search adds $0.0025.Q · 03 Is this the same as the old Sonar model? +
sonar and sonar-pro users to the Agent API's fast preset, sonar-reasoning-pro to low and sonar-deep-research to high - presets that run third-party models. perplexity/sonar is listed separately, as Perplexity's grounded search model, at a different rate card from the old Sonar row's $1/$1 plus request fees.Q · 04 Do my existing Sonar API calls still work? +
Q · 05 Are flex or priority tiers available? +
flex at 0.5x and priority at 2x the listed token prices on supported models, but the models page lists no service tiers for perplexity/sonar; an unsupported tier is ignored and the request runs at default processing.Q · 06 What is the context window? +
perplexity/sonar. This page leaves it as not listed rather than carrying over a figure from the retired Sonar API.Q · 07 How does it compare with the models Perplexity's presets use? +
$0.10/$0.50 per 1M, which Perplexity's changelog names as the model behind the low and medium presets. Presets add their own tool calls and token budgets, so compare per run, not per token.