Prompt Piggy

Value for money of AI subscriptions

How much model quality do you get for your money? The ranking combines the price in your country with the quality of the strongest included model according to Epoch AI. Below it, we compare per provider how much allowance you get per euro when you move to a bigger plan. Every input is open – if one is missing, there is no score rather than a guess.

Score = quality × 100 / monthly price in €

Ranking: quality per euro

United States: providers list prices before tax (sales tax or GST/HST varies by state/province). All prices in this ranking are therefore net – comparable with each other, but not with gross prices in other countries.

RankPlanPrice/monthQuality (ECI)Scorebest = 100
1ChatGPT GoUS$8.00plus taxGPT-5.6 LunaECI 156.5 · q = 0.92810012.91 per €100
2Claude ProUS$20.00plus taxClaude Opus 5.5ECI 168.6 · q = 1.000435.57 per €100
3Google AI ProUS$19.99plus taxGemini 4 ArgonECI 166.3 · q = 0.986435.49 per €100
4ChatGPT PlusUS$20.00plus taxGPT-6 SolECI 163.6 · q = 0.970425.40 per €100
5Claude Max 5xUS$100.00plus taxClaude Opus 5.5ECI 168.6 · q = 1.00091.11 per €100
6Google AI Ultra (5x)US$99.99plus taxGemini 4 ArgonECI 166.3 · q = 0.98691.10 per €100
7Claude Max 20xUS$200.00plus taxClaude Opus 5.5ECI 168.6 · q = 1.00040.56 per €100
8Google AI Ultra (20x)US$199.99plus taxGemini 4 ArgonECI 166.3 · q = 0.98640.55 per €100
No score – and why

These plans are sold in United States, but at least one input is missing. We do not estimate.

  • ChatGPT Pro (entry tier): model not in the Epoch index
  • ChatGPT Pro: model not in the Epoch index
  • Google AI Plus: provider names no model version
  • Microsoft 365 Personal with Copilot: provider names no model version
  • Microsoft 365 Family with Copilot: provider names no model version
  • Microsoft 365 Premium: provider names no model version
  • Perplexity Pro: provider names no model version
  • Perplexity Max: provider names no model version

Is the upgrade worth it? Allowance per euro by provider

Allowances cannot be compared across providers: “20x Pro” at Claude and “50x Plus” at Grok refer to different base amounts that the providers do not quantify. We therefore compare them only within one provider, measured against the reference plan (= 1). Above 1 means more usage per euro than the reference plan – worth it if you need it.

Allowance per € = allowance × price of the reference plan / price

Claude

PlanPrice/monthAllowanceAllowance per €reference plan = 1
Claude ProUS$20.00plus tax1×reference plan1.00×
Claude Max 5xUS$100.00plus tax5× Claude Proofficial multiplier1.00×
Claude Max 20xUS$200.00plus tax20× Claude Proofficial multiplier2.00×

Gemini

PlanPrice/monthAllowanceAllowance per €reference plan = 1
Google AI PlusUS$4.99plus tax0.5× Google AI Proofficial multiplier2.00×
Google AI ProUS$19.99plus tax1×reference plan1.00×
Google AI Ultra (5x)US$99.99plus tax5× Google AI Proofficial multiplier1.00×
Google AI Ultra (20x)US$199.99plus tax20× Google AI Proofficial multiplier2.00×

The formula

Score = quality × 100 / monthly price in €

Allowance per € = allowance × price of the reference plan / price

  • Quality: Epoch Capabilities Index (ECI) of the strongest chat model included in the plan, divided by the highest ECI of all mapped models (Claude Opus 5.5: 168.64). The strongest model scores 1.0. Models that cost extra or are only available outside chat (agents, coding) do not count.
  • Price: monthly price in the selected country incl. VAT on monthly billing (otherwise the annual plan per month), converted to euros at the ECB rate for the formula. Net in the US and Canada, see note.
  • The score (“quality per €100”) is shown relative to the best plan in the country (= 100). The allowance is not part of it.
  • Allowance (only in the per-provider comparison): usage relative to the provider's reference plan (ChatGPT Plus, Claude Pro, Google AI Pro, Le Chat Pro, Grok Plus = 1), from official numbers or multipliers such as “5x Pro” in our limits data; multipliers stated against the free tier are chained (AI Plus = 2× ÷ 4× = 0.5× AI Pro). Allowances cannot be compared across providers, because no one states how much a reference plan offers in absolute terms.

Allowance confidence levels

reference plan
The provider's reference plan, 1× by definition.
official number
Both plans state a concrete number for the same feature (e.g. messages per day).
official multiplier
The provider states a factor (“5x more usage than Pro”), not an absolute amount.
measured
Our own dated measurement with method and sample size – none yet.

What the score does not tell you

  • The ranking says nothing about how much you can use. Allowances cannot be compared across providers and therefore appear only in the per-provider comparison.
  • Many multipliers are minimums (“at least 5x”) and apply per 5-hour window; hardly any provider states weekly limits.
  • The ECI measures general capability across many benchmarks. Current frontier models are only a few percent apart and their confidence intervals overlap, so price moves the score more than quality.
  • Features such as image and video generation, storage, deep research or coding agents are not included.

Data as of

  • Prices: 30 October 2026
  • Limits: 27 October 2026
  • Model mapping: 29 October 2026
  • Quality data retrieved: 29 October 2026

AI prices and models change fast – the score is only valid for this date.

Quality data: Epoch AI (CC BY 4.0)

Epoch Capabilities Index (ECI) · eci_scores.csv

Methodology →

Which model counts

Maintained mapping with sources (data/plan-models.json). No stated version, no mapping.

PlanQuality (ECI)Which model counts
ChatGPT GoGPT-5.6 LunaECI 156.5GPT-5.6 Luna (Instant and Think)OpenAI Help: GPT-6 and other models in ChatGPT · 29 October 2026
ChatGPT PlusGPT-6 SolECI 163.6GPT-6 Sol (Thinking Medium/High)OpenAI Help: GPT-6 and other models in ChatGPT · ChatGPT Pricing · 29 October 2026
ChatGPT Pro (entry tier)GPT-6 Promodel not in the Epoch indexGPT-6 Pro + GPT-5.6 Sol ProOpenAI Help: GPT-6 and other models in ChatGPT · 29 October 2026
ChatGPT ProGPT-6 Promodel not in the Epoch indexGPT-6 ProOpenAI Help: GPT-6 and other models in ChatGPT · 29 October 2026
Claude ProClaude Opus 5.5ECI 168.6Opus: YesClaude Pricing · Claude Docs: Models overview · 29 October 2026
Claude Max 5xClaude Opus 5.5ECI 168.6Opus: Yes · Fable: 50% of weekly limitsClaude Pricing · Claude Docs: Models overview · 29 October 2026
Claude Max 20xClaude Opus 5.5ECI 168.6Opus: Yes · Fable: 50% of weekly limitsClaude Pricing · Claude Docs: Models overview · 29 October 2026
Google AI Plusno version statedAccess to our Flash modelGoogle AI subscriptions · 29 October 2026
Google AI ProGemini 4 ArgonECI 166.3Higher access to Gemini 4 ArgonGoogle AI subscriptions · 29 October 2026
Google AI Ultra (5x)Gemini 4 ArgonECI 166.3Highest access to Gemini 4 ArgonGoogle AI subscriptions · 29 October 2026
Google AI Ultra (20x)Gemini 4 ArgonECI 166.3Highest access to Gemini 4 ArgonGoogle AI subscriptions · 29 October 2026
Mistral Le Chat Pro (students)no version statedChat with our most powerful, SOTA modelsMistral AI Pricing · 29 October 2026
Mistral Le Chat Prono version statedChat with our most powerful, SOTA modelsMistral AI Pricing · 29 October 2026
Microsoft 365 Personal with Copilotno version statedCopilot Chat: Extensive usageMicrosoft Support: AI credits and limits · 29 October 2026
Microsoft 365 Family with Copilotno version statedCopilot Chat: Extensive usageMicrosoft Support: AI credits and limits · 29 October 2026
Microsoft 365 Premiumno version statedCopilot Chat: Extensive usageMicrosoft Support: AI credits and limits · 29 October 2026
Grok Plusno version statedShared AI usage poolGrok Plans · 29 October 2026
Grok PremiumGrok 4.7ECI 154.4Latest Grok modelsGrok Plans · 29 October 2026
Grok SuperGrok 4.7ECI 154.4Latest Grok modelsGrok Plans · 29 October 2026
Grok UltraGrok 4.7ECI 154.4Most powerful Chat modelsGrok Plans · 29 October 2026
Perplexity Prono version statedModel selectionPerplexity Pro · 29 October 2026
Perplexity Maxno version statedModel selectionPerplexity Pro · 29 October 2026