nvNeuralVerge

Perplexity Sonar Cost per Answer: What a Cited Answer Really Costs

Perplexity Sonar cost per answer is not one call's price. Count retries, follow-ups and checking, then compare NeuralVerge research (REST) and MCP tools.

By the NeuralVerge team · 13 min read · Updated October 2026

▸On this page

Perplexity Sonar cost per answer is not the price of one call. What you are buying is a cited answer that someone can act on, and getting one usually takes more than one call, plus some checking. This post builds a cost-per-accepted-answer model you can fill in with your own numbers and applies it to Sonar and to NeuralVerge's AI research, which runs over REST, with Search and extraction also available as MCP tools. For the feature-level comparison, read the Sonar alternative post.

The short version

  • —The verdict: cost per answer is (first attempt + retries + follow-ups) divided by the acceptance rate, plus the cost of checking. The call price is one input, and the check is often the biggest term.
  • —Who it is for: teams that ask compound or high-stakes questions, such as ownership and funding together, where a reviewer has to confirm the citations before the answer is used.
  • —What NeuralVerge charges: research is priced by depth tier, from 20 points for Lite to 400 for Ultima. Search and AI extraction are 5 points per call. The pricing page lists the rest.
  • —What this post will not do: quote Perplexity prices. They change, so take them from Perplexity's pricing page and from your own charges.

Why the price of a call misleads

Pricing pages list prices per request, per million tokens or per plan. Those are prices for inputs. Nobody's requirement is "make 10,000 calls"; it is "give the analyst a sourced answer about this company". Four things hide in the gap.

  • —Attempts. The first response is sometimes not the one you keep. The question was ambiguous, the sources were thin, or the answer wandered off the point. Each additional attempt is another charge for the same answer.
  • —Follow-ups. A single question often leads to a second: which of these is the parent company, what changed since last year, why do these two sources disagree. In a real workflow one answer is frequently a small conversation.
  • —Verification. A citation is only worth what it costs to check. If a reviewer has to open six links to confirm one sentence, that time is part of the cost of the answer, and it is usually the largest part.
  • —Rejections. Some answers are simply not usable. They are paid for and thrown away, and the money they cost has to be spread across the ones that were kept.

A comparison that ignores these four compares price lists. One that includes them compares what you will spend.

How each side charges

Per-request and token-based pricing

Search-grounded answer APIs of the Sonar kind are generally priced around what moves through the system on each call: the tokens sent to and generated by the model, and some measure of how much search context or how many searches the request used. Different model tiers carry different rates. That is a reasonable model. Cost follows volume, small requests cost less than large ones, and you can predict spend if your requests are all about the same size.

What it does not do is tie price to how much work a question deserved: the price tracks text moved, not effort spent. For similar questions that is fine. With a wide spread of difficulty, per-call cost becomes a distribution you have to measure.

For the formula below, run a representative sample and record the actual charge per call rather than estimating from token counts.

Selectable depth tiers

NeuralVerge runs a research task at a depth tier you choose for that task. Points and turnaround are as listed on the pricing page and in the documentation:

TierPoints per taskListed turnaroundDescribed as
Lite2030s to 90sLightweight and fast
Base501m to 2mEfficient for many tasks
Core1002m to 4mBalanced and strong for many tasks
Pro2003m to 7mExploratory deep search
Ultima4005m to 12mExtensive deep search

You pick the depth. A narrow lookup can go to Lite. A broad or ambiguous question can go to Pro or Ultima. For questions that only need ranked links, Search returns results in a single synchronous call at 5 points, with no research task at all. The turnaround times are the ranges listed in the documentation for each tier. We have not measured them against another provider.

Picking the finalizer separately

Depth tier controls how much searching and reasoning a task gets, but NeuralVerge also lets you choose the model that writes the final answer, independently of that tier. The settings object on a research call accepts a finalizer_model parameter alongside the depth setting, so a team can run a lighter search pass and still have a stronger model synthesize the final memo, or the reverse.

The finalizer step is what a reviewer reads. If synthesis quality drives rejections, upgrading only the finalizer is more targeted than moving the whole task up a tier.

The rest of the NeuralVerge levers

Depth and the finalizer are two levers. The others act on the checking cost (V below), such as per-claim citations and a cross-check step. The full set:

NeuralVerge advantageWhat it means for you
Per-claim citationsEvery statement in the answer links to the source behind it, so a reviewer can check one sentence without re-reading a list of links
Primary sources firstCompany and person facts come from 300+ structured sources — corporate registries, company intelligence, enrichment, professional profiles — before the open web
A dedicated cross-check stepFacts are checked against more than one source, and disagreements are surfaced instead of merged into one confident sentence
Five depth tiers per requestFrom a quick Lite lookup (30s–90s) to an exhaustive Ultima investigation (5m–12m), chosen question by question
Selectable finalizer modelChoose the model that writes the final answer independently of how deep the search goes
One platform for the whole agentResearch, Search, AI Extract and source lookups on one account with one response shape — Search, AI Extract and every data source are also MCP tools

The cost model

The formula

Here is a formula you can use for either provider, or for any other. All the variables are things you can measure in your own environment.

Cost per accepted answer = (spend on all calls + cost of checking) divided by the number of answers accepted

Written out with the pieces that matter:

  • —N is the number of distinct questions you ask in a period.
  • —A is the acceptance rate: the share of those questions that end with an answer someone actually uses, after all attempts.
  • —C1 is the average charge for a first attempt.
  • —r is the average number of extra attempts per question (retries, rephrased questions, a second run at a higher depth).
  • —Cr is the average charge for each extra attempt.
  • —f is the average number of follow-up questions per question.
  • —Cf is the average charge for each follow-up.
  • —V is the checking cost per accepted answer, as time multiplied by an hourly cost, or a flat allowance per answer.

Then Total spend = N × (C1 + r × Cr + f × Cf), and Cost per accepted answer = (Total spend + N × A × V) ÷ (N × A), which simplifies to:

(C1 + r × Cr + f × Cf) ÷ A + V

Two things stand out when you read it that way. First, acceptance rate sits in the denominator of the whole spend term. If A drops from 0.9 to 0.6, the cost of every accepted answer goes up by half, whatever the per-call price is. A cheaper call with a lower acceptance rate can lose to a dearer call with a higher one.

Second, V is added on its own and is not divided by A, because you only check answers you intend to use. It is often larger than every other term put together. If a person needs two minutes to confirm the citations behind an answer, at almost any realistic hourly cost that is more than the machine spent producing it.

Where the hidden cost sits

Retries. A retry is triggered by a bad first answer, so the way to cut retry cost is to match effort to the question, which is what a depth tier is for.

Follow-ups and rejections. Treat a question and its usual follow-up as one unit, since follow-ups are easy to forget in a budget. Rejected answers are paid for and discarded, and their cost is spread across the accepted ones through the A in the formula. That is why cost should be stated per accepted answer, never per call.

Verification. This is the largest term for most teams and the one most often left out. What drives it is how easily a reader can go from a sentence to its evidence. A list of sources at the end of an answer means opening several links to confirm one sentence; citations attached to each claim cut that to one link. For cost purposes the takeaway is narrow: measure V under each option using the same reviewer and the same questions. We are not claiming a specific reduction. If per-claim citations bring the check from three minutes to one, that saving multiplies across every accepted answer and can outweigh the difference in machine spend entirely.

Worked example: Acme Oy (Finland)

The numbers in this section are illustrative. They are chosen to show how the formula behaves, not to describe how either product performs. Replace them with your own.

Say a team researches suppliers and needs an answer for each of 1,000 companies in a month. The typical question is: "Who owns Acme Oy (Finland), and has it raised any funding in the last two years? Cite each claim."

Scenario A: a per-call model. Call the price of a call P, and do not put a real number in it: take P from the provider's pricing page for the model tier you would use, and confirm it against a few real charges. Suppose:

  • —Each question needs one call, and 20 percent need one retry: r = 0.2, Cr = P.
  • —Half of the questions prompt one follow-up: f = 0.5, Cf = P.
  • —Acceptance after all of that is 80 percent: A = 0.8.
  • —Checking takes one person 3 minutes per accepted answer, valued at V.

Then cost per accepted answer is (P + 0.2P + 0.5P) ÷ 0.8 + V, which is 2.125 × P + V. You pay for a little over two calls per accepted answer, plus the check. The point is not the 2.125. The point is that it is not 1.

Scenario B: a depth-tiered pipeline. Now the same 1,000 questions on NeuralVerge, choosing a tier per question. Suppose 700 are simple lookups that go to Base (50 points each) and 300 are compound questions that go to Core (100 points each). To keep the comparison honest, assume the same friction as before: 20 percent need a second run at the same tier, half prompt a follow-up at the same tier as the original, and acceptance is 80 percent.

  • —First attempts: 700 × 50 + 300 × 100 = 65,000 points.
  • —Retries and follow-ups add 0.7 times that: 45,500 points.
  • —Total: 110,500 points, for 1,000 × 0.8 = 800 accepted answers.

That is 138.125 points per accepted answer, before the check. The tier points are as listed on the pricing page; the routing split, retry rate, follow-up rate and acceptance are assumptions. Whether 138 points beats 2.125 × P depends on P and on what a point costs on your plan, so work it out from your own invoice rather than from a number we print.

The comparison that matters is that figure plus V against 2.125 × P plus V, with your own routing, acceptance rates and V.

REST and MCP

NeuralVerge research tasks, the depth tiers above, run over the REST API. They are not MCP tools, because a research task is asynchronous. A task is started with a POST to run-research, and the result is read from get-session-status once its status is complete. The body for the Acme Oy question at the Core tier looks like this:

{
  "instructions": "Summarize the corporate structure and ownership of Acme Oy (Finland)",
  "settings": { "country_code": "fi", "search_enabled": true, "deepsearch_model": "core" }
}

At the Core tier that task costs 100 points. Exact parameters, including finalizer_model, are in the documentation.

What your agent can use over MCP are the pieces around it: Search (run_search), AI extraction (run_extract) and every data source in the catalog, each exposed as a tool. Connecting takes one URL: add https://mcp.neuralverge.ai/mcp to an MCP client such as Claude, ChatGPT, Cursor or VS Code, sign in through the browser and click Allow. There is no key to paste and no server to host. The MCP server for Claude walkthrough covers setup.

MCP does not change the cost model. A tool call is priced the same as the equivalent REST call and returns the same response shape, so there is no separate MCP rate to add to the formula. What it can change is the attempts and follow-ups terms: an agent that calls a source lookup mid-task for a single fact, instead of starting a research task, avoids paying for depth the question did not need.

Measure it on your own questions

A short procedure gets you real numbers in an afternoon.

  1. —Build a question set. Take 30 to 50 questions from your actual workload, including a few of the awkward ones. Do not curate it to flatter any provider.
  2. —Fix the rules for acceptance. Decide in advance what counts as a usable answer: correct entity, all the requested fields present, each claim traceable to a source. Have the person who would use the answers do the marking.
  3. —Run each option the way you would in production. That includes your normal retry policy and the follow-ups a user would realistically ask.
  4. —Record every charge. For each option, read the charge for each request or task from its usage records rather than computing it from a price sheet.
  5. —Time the checking. For every accepted answer, have the reviewer confirm its citations and record the minutes. Convert to money with an hourly cost that suits you.
  6. —Compute the formula. Work out A, r, f, C1, Cr, Cf and V from the records, then the cost per accepted answer for each option.

If two options come out close, the tiebreaker is usually how easy the citations are to verify, since that affects V for every answer you keep.

Routing between the two

Many systems use both, routing by question type. A fast, single-pass call fits chat-facing questions and high volumes of small, similar ones, where retries and follow-ups are near zero. A multi-step pipeline with selectable depth fits compound questions and work where the check is expensive. Sonar and similar APIs against NeuralVerge, dimension by dimension:

DimensionPerplexity Sonar & other LLM search APIsNeuralVerge
What you get backRanked web results, page content, or a generated answer over them — depending on the product and modeA finished, synthesized answer with inline citations — or a structured record straight from a named source
Where facts come fromMostly a general web index or the open webStructured sources first — corporate registries, company intelligence, contact enrichment, professional profiles — with the open web as fallback
Company & person recordsWhatever web pages happen to say about the company or personDirect lookups across 300+ sources: register records, officers and filings, funding rounds and investors, work email, phone, professional profiles
CitationsVaries — check whether sources are attached per claim or as one list per responsePer claim — every statement in the answer links back to its source
Cross-checkingUsually left to your agent or your promptA dedicated step checks facts against more than one source and surfaces disagreements instead of smoothing them over
Depth controlVaries by product — modes, processors or model tiersFive tiers chosen per request, from Lite (30s–90s) to Ultima (5m–12m)
Synthesis modelVaries by productSelectable per task via finalizer_model, independently of depth tier
Agent accessREST; many also ship an MCP serverAI research over REST; Search, AI Extract and every data source also as tools on a hosted MCP server
Best fitFast single-fact questions, chat answers, broad web topicsCompany and person questions that need primary sources and a citation trail — account research, KYB, due diligence
The middle column describes how this category of tools typically works; individual products differ, so check each vendor's current documentation.

How to decide

  • —Is your workload mostly small, similar questions answered correctly the first time? Then retries and follow-ups are near zero and the call price dominates. Measure that before assuming you need more.
  • —Are your questions compound, such as ownership and funding together? One search pass has to cover meaningfully different ground, which is where acceptance tends to drop and retries rise.
  • —Is the check the expensive part? Due diligence, supplier vetting and background research are cases where per-claim citations are meant to shorten review. Time the check under each option.
  • —Does question difficulty vary widely? Choosing a tier per task lets you spend more only where a question warrants it. Without per-request depth, easy and hard questions get the same amount of work.
  • —What is the retry rate on your questions? Measure it. It multiplies your spend directly.
  • —Can your team measure acceptance? If nobody will mark answers as accepted or not, no cost model will be better than the price list.

Frequently asked questions

Why not just compare the list price of one call?

Because one call is rarely one answer. Retries, follow-up questions and the time a person spends checking citations all belong to the cost of an answer you can actually use. A cheap call that needs three attempts and a manual check can cost more than a pricier call that is accepted the first time.

How do I estimate acceptance rate without guessing?

Measure it. Take a sample of real questions, run them through each option, and have the person who would use the answer mark each one accepted or not. Acceptance rate is the accepted count divided by the total. A sample of a few dozen questions from your own workload tells you more than any published figure, including ours.

Where do I get the Perplexity Sonar price for the formula?

From Perplexity's own pricing page for the model tier you would use, confirmed against a few real charges in your usage records. This post does not quote Sonar prices because they change. Run a sample of your questions and record the actual charge per call rather than estimating from token counts.

Can I use both in one system?

Yes. Nothing stops you from sending quick, single-fact questions to a fast search-grounded call and sending compound or higher-stakes questions to a deeper pipeline. Routing by question type, and measuring cost per accepted answer for each route, is a sensible way to run both.

Can I use NeuralVerge over MCP, or only through the API?

Both, with one exception. Search, AI extraction and every data source are MCP tools on the hosted server at https://mcp.neuralverge.ai/mcp, priced the same as the REST call and returning the same response shape. AI research tasks, the depth tiers, are called over the REST API and are not an MCP tool.

About NeuralVerge

Give your agents structured, cited, real-world data from 300+ sources and the open web — through one API or MCP server.

AI Deep Research on the NeuralVerge blog.