Perplexity Model Council runs on Pro through Computer, paid in credits, without the $200 Max plan.

Max adds capped web and iOS runs, and the cap has no published number.

A council of three models built on the Agent API costs $0.0802 per run in tokens. Checked October 30, 2026.

Three ways to run Model Council: plans, where each runs, limits, and price.

Path

Plans

Where it runs

Models per run

Limit

Price

Web and iOS

Max, Enterprise Max

Web, iOS app 26.16.0 or later

3

Monthly cap, number not published

$200 per month or $2,000 per year †

Computer

All paid plans

Computer

2 to 8

Credits, 100 credits for $1, no credit price per run published

Pro: $17 a month billed yearly, no monthly credits †; Max: 10,000 credits a month †

API council

Any API account

Agent API

3 and a synthesizer

Tokens

$0.0802 a run, derived

† Reported by Perplexity pages, not verified by LLM Waves Research. The API council price is arithmetic on published token rates.

Web and iOS Council needs Max or Enterprise Max and stops at a monthly cap that Perplexity does not number.

Council in Computer runs on all paid plans and draws on credits at 100 credits for $1.

An API council of three models and a synthesizer costs $0.0802 per run in tokens.

Where these numbers come from

Every plan, credit, and token figure comes from Perplexity's help pages, pricing page, and Agent API model list, read on 2026-10-30. LLM Waves Research did not run Model Council or any model.

The API council price is calculated from published token rates, and the formula is in the data file with each source URL and retrieval date.

Method notes are on the LLM Waves methodology page.

This comparison ranks the three paths on plan fee first, cost per run second, and limits third.

Plan fees set the gap before any credit is spent.

Pro at $17 a month for 12 months is $204, and $2,000 minus $204 is $1,796.

Max costs 9.8 times the Pro fee for the year.

Max and Enterprise Max get web and iOS Council under a cap with no published number

The Perplexity Model Council help page lists web and iOS Council for Max and Enterprise Max, and lists that mode as not available on Free, Pro, Education Pro, or Enterprise Pro.

The Android and desktop apps do not have the standalone mode yet.

Web Council always runs three models. The default lineup is one model each from Anthropic, OpenAI, and Google, and a synthesizer model writes the combined answer.

Perplexity's model list page, as of October 27, 2026, adds GPT-6 Astra, Gemini 4 Argon, Claude Opus 5.5 and Claude Fable 5.1 on Max, and states that Council has its own model selection.

Perplexity's own models are on the Perplexity creator page.

A monthly cap limits web and iOS queries, and the help page gives no number.

The Perplexity pricing page says Max includes "several Model Council runs per month" and lists Max at $167 a month when paid yearly.

The help page lists Max at $200 per month or $2,000 per year, and $2,000 ÷ 12 is $166.67.

Pro runs Council in Computer, and credits pay for it

The Perplexity announcement of Council in Computer, dated July 28, 2026, lists it for all Max users and for Pro users in Computer, with two to eight models.

The help page lists Council in Computer on all paid plans and lists the monthly cap as not applying there, since Computer runs spend credits.

The pricing page FAQ lists model council under Pro. The help page lists web and iOS Council as not on Pro.

Both statements fit one reading: Pro runs Council in Computer and not in the web and iOS mode.

Credits by plan, in dollars

The Perplexity credits page sets 100 credits at $1. Unused monthly credits do not roll over, and bonus credits expire in about 30 days.

  • Pro: no monthly credits. New subscribers get 4,000 bonus credits one time, worth $40.

  • Max: 10,000 credits a month, worth $100. New subscribers get 35,000 bonus credits one time, worth $350.

  • Enterprise Pro: 500 credits a month, worth $5.

  • Enterprise Max: 15,000 credits a month, worth $150.

The Max credits are worth 50% of the $200 monthly price and 60% of the $166.67 yearly rate.

Monthly credits are where the plans differ most.

A Pro reader has $0 of monthly credits and relies on the welcome bonus or on bought credits.

Pro weekly usage and Light effort

Paid consumer Pro plans include a weekly amount of Computer usage that resets each week and is not a credit. With no credits, Computer runs tasks on Light effort, and other effort levels need credits.

The credits page says the weekly usage is rolling out gradually. Perplexity does not say whether a council runs on Light effort, so that point is unconfirmed.

The Pro welcome credits cover 4 to 11 research reports at most

Perplexity publishes no credit price for a Council run. The credits page lists task anchors for Computer: 15 to 70 credits for a quick fact and 350 to 950 credits for a multi-step research report.

The help page says a council costs more than a single model session, and each added model raises the cost.

The 4,000 welcome credits buy 4.2 reports at 950 credits each and 11.4 at 350 credits each. Rounded down, that is 4 to 11 reports, or 57 to 266 quick facts at 15 to 70 credits each.

One month of Max credits buys 10 to 28 reports by the same steps. A council costs more than one model, so these counts are upper bounds and not Council prices.

Max monthly credits cover 2.5 times the reports of the Pro welcome credits.

A council built on the Agent API costs $0.0802 per run

The Agent API model list prices Claude Opus 5.5 at $4 per 1M input tokens and $20 per 1M output tokens.

GPT-6.1 Sol and Claude Sonnet 5.5 are each $2 and $10. Gemini 4 Argon is $2 and $10 as an introductory rate through November 30, 2026, and the page lists no rate after that date.

Take a council of three models and a synthesizer. Each model reads 1,500 tokens and writes 1,200 tokens, reasoning tokens included. The synthesizer reads 5,100 tokens and writes 1,000 tokens. Search and tool fees are left out.

Claude Opus 5.5 reads 1,500 tokens at $4 per 1M, which is $0.0060. It writes 1,200 tokens at $20 per 1M, which is $0.0240. One answer costs $0.0300.

GPT-6.1 Sol and Gemini 4 Argon each cost $0.0150 for one answer.

Claude Sonnet 5.5 writes the combined answer for $0.0202.

The four add up to $0.0802 per run. The Perplexity API pricing guide covers the presets.

The cost per run climbs with every model added.

Each added model costs $0.0150 for its answer and $0.0024 more for the synthesizer to read it. Swapping one model for Claude Opus 5.5 lifts three models from $0.0652 to $0.0802.

Computer bills in credits, so these token costs do not predict a Computer bill.

Max pays for itself at 2,078 council runs a month, on token cost alone

Max costs $200 a month, and $200 ÷ $0.0802 is 2,494 runs. Max paid yearly costs $166.67 a month, and $166.67 ÷ $0.0802 is 2,078 runs.

Below those counts, an API council costs less than the Max fee for council runs alone.

The word "several" on the pricing page suggests a cap far below 2,078, but the page gives no number, so the Max fee per web run cannot be priced.

Past 2,078 runs a month, the Max fee costs less than the same runs on the API.

What we did not measure

This article did not run a latency test, a throughput test, or an output-quality benchmark against any model discussed. No Council run was made on any account.

The credit price of a Council run, the number in the monthly cap, and the model list for Council in Computer are not published, so none appears here.

Council on Light effort and Council in Computer on Android are unconfirmed.

Which path fits which reader

A Pro reader trying Council: Top pick, Computer on Pro. The plan fee is $204 for the year against $2,000 for Max, and the 4,000 welcome credits cover at most 11 research reports.

If Council does not run on Light effort or one run uses most of the welcome credits, this pick is wrong for you and Max, or the API council fits better.

A heavy web or iOS user: Insufficient data. The cap has no published number, so the Max fee per run cannot be computed. A reader who needs a set number of runs a month has to read the cap on a Max account before paying $200.

A developer: Situational, the API council at $0.0802 a run. It fits only when Council has to run from code, since the Agent API lists no Council endpoint. If you need the exact synthesizer, this pick is wrong for you, because the help page does not name it.

An enterprise team: Situational, Enterprise Max. It carries web and iOS Council and 15,000 monthly credits worth $150. If the team needs only Computer runs, the web and iOS mode adds nothing for it, and Enterprise Pro carries Computer with 500 monthly credits worth $5.

Does Perplexity Pro have Model Council?

Yes, through Computer. Perplexity lists Council in Computer on all paid plans, paid in credits.

Pro has no web or iOS Council and no monthly credits, only the welcome bonus.

Is Perplexity Model Council free?

No. The help page lists the standalone mode as not available on Free, and Computer is a paid plan feature.

The pricing page lists Pro at $17 a month when billed yearly.

Is there a Perplexity Model Council API?

No. The Agent API overview lists no Council endpoint, parameter, or preset. It documents Model Fallback, which takes several models for automatic failover.

A council built from separate calls is the only API route.

Does Perplexity Model Council work on Android?

The help page lists the standalone mode as not yet available in the Android app or the desktop apps.

It does not say whether Council in Computer runs on Android, so that part is unconfirmed.

How many models does Perplexity Model Council run?

Web and iOS runs always use three models. Computer takes two to eight models, or the default lineup.

Read the cap on a Max account before paying for it

A reader choosing between $200 for Max and credits on Pro needs two numbers that Perplexity has not published: the monthly cap and the credit price of one council run.

Until those two numbers appear, a first run on Pro costs less to try than a $2,000 year of Max.