Claude and Codex pricing in AED and USD: plans vs API value

A subscription’s API value is what the same month of usage would cost at the vendor’s pay-per-token rates. Measured that way, Claude’s $200 Ultra plan buys about $11,700 of API-priced usage a month, about as much as OpenAI’s $500 ChatGPT Pro tier at $10,400–14,200. Every plan below is a deep discount on the API, if you use most of its limit.

The figures are estimates built from published measurements, adjusted for OpenAI’s 29 October change, in US dollars and UAE dirhams. Prices checked 1 November 2026. For how I split the actual work between Claude Code, Codex and the other tools, see the toolbench.

How much does Claude cost? Pro, Max and Ultra in AED and USD

Claude has three paid plans for individuals. Pro costs $20 a month (AED 73), or $17 a month billed yearly. Max costs $100 (AED 367) and gives five times Pro’s usage in every five-hour window and every week. Ultra costs $200 (AED 735): ten times Pro’s weekly usage, with no five-hour limit for verified accounts. Claude Code is included from Pro upwards; the free plan does not include it. Claude plans and pricing.

Max and Ultra add a monthly API credit for the Claude Platform ($100 on Max, $200 on Ultra) and let Claude Fable 5.1 use up to half of the weekly limit. On Pro, Fable is paid from usage credits.

AED figures convert at the fixed peg of 3.6725 dirhams to the dollar and exclude VAT. Whether 5% UAE VAT appears on your bill depends on how the vendor invoices you, and your bank may add a foreign-currency fee, so check one statement before budgeting.

How much is ChatGPT Pro? Plus and Pro $100, $200 and $500 for Codex

Codex runs on ChatGPT plans. Plus costs $20 a month (AED 73). Pro comes in three tiers: $100 (AED 367), $200 (AED 735) and $500 (AED 1,836). Free ChatGPT accounts get limited Codex use, Plus gets standard use and Pro gets expanded use. Pro tiers also include monthly API credits that rise with the tier. ChatGPT plans and pricing.

On 29 October 2026 OpenAI dropped a planned cut to Pro usage and doubled it instead. Pro $100 and Pro $500 now include twice their previous Work and Codex allowance; on Pro $200 only the lower of its two allowances doubled. About ChatGPT Pro tiers; BleepingComputer’s report.

Only Pro $500 includes Ultrafast, a lower-latency mode for GPT-6 Astra and GPT-6.1 Sol that draws included usage at eight times the Standard rate. On Plus, OpenAI publishes message ranges per five-hour window: 5–45 for GPT-6 Astra, 15–160 for GPT-6.1 Sol and 350–3,000 for GPT-6 Luna. Managing usage with GPT-6 Astra.

How many tokens does Claude Pro (and Max, Ultra) actually give you?

Neither vendor publishes a token allowance. Anthropic describes limits as multiples of Pro; OpenAI gives message ranges. The workable stand-in is measured usage priced at API rates. SemiAnalysis measured Claude Pro at about $1,180 of API-priced usage a month at full use, roughly 59 times its price. It also found that Claude plans give about 2.8 times as many tokens as the matching ChatGPT plans.

In tokens at Opus 5.5 rates, Pro’s $1,180 is about 160 million a month with an 80/20 input and output mix and no caching, or about 430 million in a cache-heavy agent loop. Max works out at about 800 million to 2.1 billion, and Ultra at about 1.6 billion to 4.3 billion. These are estimates; neither vendor publishes a token allowance.

Plan price against estimated API value per month at full use, checked 1 November 2026
Plan	Price	AED, ex-VAT	API value per month	Multiple	Basis
Claude Pro	$20	AED 73	$1,180	59x	SemiAnalysis, measured
Claude Max	$100	AED 367	$5,730	57x	SemiAnalysis, measured
Claude Ultra	$200	AED 735	$11,730	59x	SemiAnalysis, measured as Max 20x
ChatGPT Plus	$20	AED 73	$210–285	11–14x	SemiAnalysis and InventiveHQ, measured
ChatGPT Pro $100	$100	AED 367	$2,100–3,300	21–33x	Both sources, doubled on 29 October
ChatGPT Pro $200	$200	AED 735	$4,200–5,800	21–29x	Both sources, lower allowance doubled
ChatGPT Pro $500	$500	AED 1,836	$10,400–14,200	21–28x	Projected, then doubled

Read each multiple as a ceiling. It assumes you use the whole limit every week of the month. At a quarter of the limit, Claude Max is worth about $1,430 at API rates, still 14 times its price.

Why do Fable, Astra, Ultrafast and sub-agents drain a plan faster?

A plan limit is a compute budget. A model that costs more per token at API rates spends the same allowance faster. Claude Fable 5.1 lists at $10 per million input tokens and $50 per million output, two and a half times Claude Opus 5.5. GPT-6 Astra lists at the same $10 and $50, and Ultrafast multiplies its draw by eight. Higher effort settings, such as Fable 5.1 on high, write more reasoning tokens, which bill as output.

Sub-agents multiply the job itself. Each one starts its own context. If a lead agent hands four sub-agents the same 50,000-token brief and file set, that step reads about 250,000 input tokens instead of 50,000. Cached input softens it: Opus 5.5 charges $0.20 per million for cached reads against $4 for fresh input, but only for an identical prefix.

Same job, same token mix (15% fresh input, 75% cached input, 10% output), priced at API list rates; rates are per million input and output tokens. Ultrafast is Astra at the same API rates with OpenAI’s 8x plan-usage rate applied. The bars show relative speed only, and the clock stops at 40 seconds.

The meters compare models across two vendors on one scale, so they show API cost, which is a stand-in for plan drain. Each vendor weights its own limits, and the published quota facts sit next to each meter.

Is Claude Max worth it? The break-even against the API

At Opus 5.5 API rates, a million tokens split 80% input and 20% output, with no caching, costs about $7.20. So $20 buys about 2.8 million tokens at the API, $100 about 14 million and $200 about 28 million. Below those monthly totals the API is cheaper; above them the plan wins. These are estimates for that mix, and caching moves them a long way in the plan’s favour.

Agent work clears those totals quickly: one long Claude Code session can read millions of tokens of files and context. If you reach Pro’s five-hour limit several times a week, Max at AED 367 a month is the cheaper way to keep working. Ultra makes sense when the five-hour window is what stops you: verified accounts have none, and its weekly allowance is twice Max’s. To budget a whole agent workflow in dirhams, the AED planning worksheet covers the costs around the model.

Is ChatGPT Pro worth it for Codex?

After the doubling, the three Pro tiers land at a similar multiple, roughly 21–33 times their price. So the right tier is the smallest one whose limit you actually reach. Pro $100 at AED 367 covers most heavy individual Codex use. Pro $500 pays off if you need Ultrafast or exhaust Pro $200 every week, and Ultrafast spends its included usage eight times as fast.

Claude Max vs ChatGPT Pro: which gives more API value?

At every price point Claude’s plan buys more API-priced usage. At $20 the gap is about 4–5.6 times ($1,180 against $210–285). At $100 it is about 1.7–2.7 times ($5,730 against $2,100–3,300), and at $200 about 2–2.8 times ($11,730 against $4,200–5,800). Claude Ultra at $200 sits within ChatGPT Pro $500’s range.

Part of that gap is the price per token. Opus 5.5 lists at twice GPT-6.1 Sol, so a dollar of Claude API value buys fewer tokens than a dollar of OpenAI value. Which model finishes your work in fewer tokens is a separate question; the toolbench covers how I route work between Claude Code and Codex.

How much does a Claude token cost? API prices compared

API list prices per million tokens, checked 1 November 2026
Model	Input	Cached input	Output
Claude Fable 5.1	$10	$0.25	$50
Claude Opus 5.5	$4	$0.20	$20
Claude Sonnet 5.5	$2	$0.10	$10
GPT-6 Astra	$10	$1	$50
GPT-6.1 Sol	$2	$0.10	$10
GPT-6 Sol	$2	$0.20	$10
GPT-6.1 Terra	$1	$0.05	$5
GPT-6 Luna	$0.10	$0.01	$0.50

Opus 5.5’s fast mode costs $8 input and $40 output, and GPT-6 Astra’s long-context rate is $20 and $75. Sources: Claude API pricing and OpenAI API pricing. In dirhams, an Opus 5.5 million-token mix at $7.20 is about AED 26.

Work out your own number

Pick a plan and the share of its weekly limit you use in a typical month. The value uses the middle of each estimate range.

How I calculated it

The Claude values come from SemiAnalysis’s report of 5 October 2026, Anthropic subscriptions offer 5x, which is paywalled; the figures here are as quoted by Mehmet Özel on Medium on 8 October: Pro $1,178, Max $5,725 and the $200 tier $11,726. SemiAnalysis measured that tier when it was called Max 20x. The pricing page now describes Ultra as ten times Pro’s weekly usage with no five-hour limit, so its value may differ.

The ChatGPT values combine SemiAnalysis (Plus $211, Pro $100 $1,055, Pro $200 $2,084) with InventiveHQ’s measured Codex usage of 4 October (Plus about $285, Pro $100 $1,440–1,640, Pro $200 $2,800–2,900, Pro $500 about $7,100 projected). Both predate OpenAI’s 29 October change, so I doubled the Pro rows as OpenAI’s help page describes, applying only the lower-allowance doubling to Pro $200. SemiAnalysis has no Pro $500 figure; its low end scales the Pro $200 figure by InventiveHQ’s ratio between the two tiers, about 2.5, before doubling.

The meters price one job at list rates with a cache-heavy agent mix: 15% fresh input, 75% cached input and 10% output. The calculator multiplies the middle of each range by the share you enter. Dirhams use the 3.6725 peg and exclude VAT. A post on X from 5 August 2026 claimed 81 times API value from four Max accounts. It gave no method and predates Opus 5.5’s price cut, so I leave it out of the numbers.

FAQ: is Claude Code or Codex free, and when do limits reset?

Is Claude Code free?

No. Claude Code comes with the Pro, Max and Ultra plans, from $20 (AED 73) a month, or with an API key billed per token. The free Claude plan does not include it.

Is Codex free?

Free ChatGPT accounts get limited Codex use. Plus at $20 (AED 73) gives standard use, and the Pro tiers from $100 give expanded use, now doubled on Pro $100 and Pro $500.

Does Claude Pro have a limit, and when does it reset?

Yes, two. A session limit starts with your first message and resets in full five hours later. A weekly limit resets at a fixed day and time assigned to your account, shown under Settings, Usage.

How much is ChatGPT Plus in the UAE?

It is $20 a month, AED 73 at the dirham peg before VAT. Some UAE price pages quote AED 79–85, which suggests VAT and card fees on top.

Is Claude better than ChatGPT for coding?

At the same price, Claude’s plans buy more API-priced usage: about 1.7–2.8 times at $100 and $200. Which writes better code for you depends on the task; I use both, and the toolbench note explains the split.

Suhail Mohebi · published 1 November 2026 · the cover is a generated illustration