# Which AI Subscription Gives the Most AI per Dollar? (Nov 2026)

Canonical: https://careinstitute.ai/blog/ai-subscription-value-per-dollar/
Published: 2026-11-19
Author: Michael Turon
Code licence: Apache 2.0 (https://github.com/CAREInst)

## In short

As of 19 November 2026, among plans with independent measurements, Claude plans buy the most usage per dollar. Claude Pro ($20) returns about 35× its price in API-priced usage and Claude Ultra ($200) about 34×. ChatGPT Pro $200 returns about 16×, ChatGPT Plus about 10×, and GitHub Copilot 1.5 to 2×. These are ceilings: each figure assumes the full allowance is used every week on agentic coding. Google does not publish its allowance, so its plans are not ranked.

Claude Ultra ($200) returns about 34× and, unlike Claude Pro, includes Fable 5.2, Anthropic's top-scoring model. GitHub Copilot returns 1.5 to 2× because its credits are spent at API list prices. We could not rank Google's AI plans: Google does not publish the size of its coding allowance, and we found no independent measurement.

API-priced usage is what the same work would cost at the vendor's API list prices. Every figure is a ceiling for heavy users: it assumes you use the full allowance every week on agentic coding work. The tables link each number to its source.

CARE Institute works on where an AI agent may send each request: [classify it, send it only to models cleared for it, and pick the best of those on cost and speed](/about/). A subscription is one of those destinations, and only for requests your rules clear for the cloud. Work that must stay on hardware you own needs a different setup; see [our local agent build](/blog/local-ai-agent-audit-trail/). Most guides compare plans on sticker price. We wanted a comparison on what each dollar buys, scoped and sourced, that others can check and rerun.

*Data as of 19 November 2026. Next review: February 2027.*

![Bar chart of API-priced usage per $1 of subscription. Claude Pro 35×, Claude Ultra 34×, Claude Team Premium 25×, ChatGPT Pro $200 16×, ChatGPT Pro $100 14× and Pro $500 14× (both measured before an October limit increase), ChatGPT Plus 10×, Copilot Max 2.0×, Copilot Pro+ 1.8×, Copilot Pro 1.5×.](figures/fig1-usage-per-dollar.png "API-priced usage per $1 of subscription. Scope: agentic coding, 89 to 92% of input served from cache, full monthly use, API list prices. Measured 9 to 16 Nov 2026, except ChatGPT Pro $100 and $500 (Oct 2026).")

*Sources: [Wokey, Claude](https://wokey.ai/claude-usage-limits); [Wokey, Codex](https://wokey.ai/codex-usage-limits); [Inventive HQ](https://inventivehq.com/blog/codex-subscription-usage-measured); [ContextBolt](https://contextbolt.com/blog/codex-usage-limits/); [SemiAnalysis via explainx](https://www.explainx.ai/blog/semianalysis-claude-vs-chatgpt-subscription-api-value-limit-testing-2026); [GitHub Docs](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-individuals). Data: [CSV](https://careinstitute.ai/blog/ai-subscription-value-per-dollar/data/subscription-value-hard-numbers-2026-11-19.csv).*

**Key findings**

- At $20, Claude Pro is the best measured value: about $706 of API-priced usage per month, or 35× its price. Its best model is Anthropic's Opus 5.5, which scores 59.1 on the Artificial Analysis Intelligence Index, an independent average of benchmark scores. ([Wokey, measured 9 to 16 Nov 2026](https://wokey.ai/claude-usage-limits))
- At $200, Claude Ultra is the best measured value and the only plan at that price with Fable 5.2: about $6,830 per month (34×). Up to half of the weekly limit can run on Fable 5.2 (index 64.6), and the rest falls back to Opus 5.5. ([Wokey](https://wokey.ai/claude-usage-limits); [Anthropic Help Center](https://support.claude.com/en/articles/15424964-claude-fable-models-on-your-plan))
- ChatGPT Pro $200 buys about the same volume as a $125 Claude Team Premium seat: about $3,191 vs $3,108 per month. For the same $200, Claude Ultra buys about twice as much. ([Wokey tracker for Codex, OpenAI's coding agent](https://wokey.ai/codex-usage-limits))
- Copilot is priced close to the API. Its plans include 1,500 to 20,000 credits at $0.01 each, spent at list prices, which works out to 1.5 to 2× the plan price. Copilot gets you many models, with little volume on top of what you pay. ([GitHub Docs](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-individuals))
- The effort setting changes the cost a lot. Moving Fable 5.2 from "high" to "max" effort adds 2.7 index points for 2.9× the cost per task. ([Artificial Analysis](https://artificialanalysis.ai/models/claude-fable-5-2))

## How do we score AI subscription value per dollar?

We combined three numbers for each plan.

1. Access: API-priced usage per $1. Independent trackers ran real agentic coding sessions until each plan hit its limit, then priced the tokens at the vendor's own API list prices. We divide that monthly dollar figure by the plan price. A result of 10× means a fully used $20 plan would cost $200 through the API.
2. Intelligence ceiling: the best model you can use *inside* the allowance, scored on the [Artificial Analysis Intelligence Index v5.0.1](https://artificialanalysis.ai/models). The top score in our data is Fable 5.2 at 64.6.
3. Build potential: total monthly usage in dollars. Two plans can have the same ratio and very different volume.

From these we compute one number, the **frontier value score**:

> [!NOTE] Frontier value score
> (API-priced usage ÷ price) × (intelligence index ÷ 64.6)

A score of 30 means each $1 of subscription buys about $30 of usage at Fable-5.2-level intelligence.

We rank only plans with hard numbers: a current third-party measurement, a measurement taken before a known limit change (marked as such), or a dollar credit stated by the vendor. Plans with only estimates are listed separately under [What we don't know yet](#what-we-dont-know-yet).

## What does higher effort cost?

Most frontier models let you choose how long they "think". More effort uses more tokens, so it costs more through the API and drains a subscription allowance faster.

![Line chart of intelligence index against cost per benchmark task for six models at each effort level. Fable 5.2 rises from 53.9 at low effort ($3.10 per task) to 64.6 at max ($26.54). GPT-6.1 Sol reaches 54.6 for $1.27. Gemini 4 Neon reaches 61.4 for $6.45. Gemini 4 Carbon reaches 60.0 for $7.76. Opus 5.5 reaches 59.1 for $10.04.](figures/fig2-effort-curves.png "Intelligence against cost per benchmark task, by effort level. Scope: Artificial Analysis Intelligence Index v5.0.1; cost of the full evaluation at list API prices divided by tasks; accessed 19 Nov 2026.")

| Model and effort | Intelligence index | Cost per task (API prices) | Source |
|---|---:|---:|---|
| Fable 5.2, low | 53.9 | $3.10 | [AA](https://artificialanalysis.ai/models/claude-fable-5-2) |
| Fable 5.2, high | 61.8 | $9.08 | [AA](https://artificialanalysis.ai/models/claude-fable-5-2) |
| Fable 5.2, max | 64.6 | $26.54 | [AA](https://artificialanalysis.ai/models/claude-fable-5-2) |
| Gemini 4 Neon, xhigh | 61.4 | $6.45 | [AA](https://artificialanalysis.ai/models/gemini-4-neon) |
| Gemini 4 Carbon, high | 60.0 | $7.76 | [AA](https://artificialanalysis.ai/models/gemini-4-carbon) |
| Opus 5.5, max | 59.1 | $10.04 | [AA](https://artificialanalysis.ai/models/claude-opus-5-5) |
| GPT-6 Astra, max | 56.9 | $6.25 | [AA](https://artificialanalysis.ai/models/gpt-6-astra) |
| GPT-6.1 Sol, max | 54.6 | $1.27 | [AA](https://artificialanalysis.ai/models/gpt-6-1-sol) |
{caption="Intelligence index and cost per task by effort level"}

*Scope: Artificial Analysis Intelligence Index v5.0.1. Cost is the price of running AA's full evaluation at list API prices, divided by the number of tasks. Accessed 19 Nov 2026. Full table, including Muse, Grok, Kimi and GLM: [effort-curves-2026-11-19.csv](https://careinstitute.ai/blog/ai-subscription-value-per-dollar/data/effort-curves-2026-11-19.csv).*

Each step up in effort buys fewer points for more tokens. On a plan with a weekly limit, using max effort on every request uses up the allowance several times faster than high effort. Our inference, not a measurement: for long agent runs, "high" is usually the better trade, and "max" is worth saving for the hardest steps.

## How much API usage does each AI subscription buy?

Every plan in our data buys more than its price at API rates, but the subsidy varies by more than 20×, from 1.5× for Copilot Pro to 35× for Claude Pro.

![Log-log scatter of monthly price against API-priced usage. Claude Pro $20 buys $706, Claude Team Premium $125 buys $3,108, Claude Ultra $200 buys $6,830. ChatGPT Plus $20 buys $202, Pro $100 $1,450, Pro $200 $3,191, Pro $500 $6,955. Copilot Pro $10 buys $15, Pro+ $39 buys $70, Max $100 buys $200. Dashed lines mark 1×, 10× and 30× the price.](figures/fig3-price-vs-usage.png "Monthly price against API-priced usage. Scope: full use of each plan's allowance at the vendor's API list rates; agentic coding with heavy caching; measured Oct to Nov 2026; as of 19 Nov 2026.")

| Plan | Price / month | API-priced usage / month | Per $1 | Evidence | Source |
|---|---:|---:|---:|---|---|
| Claude Pro | $20 | $706 (other test: $1,178) | 35× | Measured, 9–16 Nov 2026 | [Wokey](https://wokey.ai/claude-usage-limits); [SemiAnalysis via explainx](https://www.explainx.ai/blog/semianalysis-claude-vs-chatgpt-subscription-api-value-limit-testing-2026) |
| Claude Ultra | $200 | $6,830 | 34× | Measured, 9–16 Nov 2026 | [Wokey](https://wokey.ai/claude-usage-limits) |
| Claude Team Premium seat | $125 | $3,108 (range $2,856–3,319) | 25× | Measured, 181 accounts, 9–16 Nov 2026 | [Wokey](https://wokey.ai/claude-usage-limits) |
| ChatGPT Pro $200 | $200 | $3,191 | 16× | Measured, 9–16 Nov 2026 | [Wokey](https://wokey.ai/codex-usage-limits); [Inventive HQ](https://inventivehq.com/blog/codex-subscription-usage-measured) |
| ChatGPT Pro $100 | $100 | $1,450 (range $1,055–1,650) | 14× | Measured in Oct, before the reported 29 Oct increase | [Inventive HQ](https://inventivehq.com/blog/codex-subscription-usage-measured); [ContextBolt](https://contextbolt.com/blog/codex-usage-limits/) |
| ChatGPT Pro $500 | $500 | $6,955 | 14× | Measured in Oct, before the reported 29 Oct increase | [SemiAnalysis via explainx](https://www.explainx.ai/blog/semianalysis-claude-vs-chatgpt-subscription-api-value-limit-testing-2026) |
| ChatGPT Plus | $20 | $202 (range $162–211) | 10× | Measured, 9–16 Nov 2026 | [Wokey](https://wokey.ai/codex-usage-limits) |
| Copilot Max | $100 | $200 | 2.0× | Vendor credit: 20,000 × $0.01 | [GitHub Docs](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-individuals) |
| Copilot Pro+ | $39 | $70 | 1.8× | Vendor credit: 7,000 × $0.01 | [GitHub Docs](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-individuals) |
| Copilot Pro | $10 | $15 | 1.5× | Vendor credit: 1,500 × $0.01 | [GitHub Docs](https://docs.github.com/en/copilot/about-github-copilot/plans-for-github-copilot) |
{caption="API-priced usage per month for each measured AI subscription, November 2026"}

*Scope: monthly figures assume the full allowance is used every week. Measurements are agentic coding sessions with 89 to 92% of input served from cache, priced at each vendor's API list rates. Claude figures are mostly Opus 5.5; OpenAI figures are mostly GPT-6.x. Download: [subscription-value-hard-numbers-2026-11-19.csv](https://careinstitute.ai/blog/ai-subscription-value-per-dollar/data/subscription-value-hard-numbers-2026-11-19.csv).*

The plans split into two groups. Claude and ChatGPT sell a block of usage per 5 hours or per week, and at API rates those blocks are worth 10 to 35× the price. Copilot sells dollar credits that are spent at list prices, so its plans sit close to the API, and the small margin above 1× is the whole subsidy.

## Which models can each plan use?

Usage on a weaker model is worth less than the same usage on the strongest one. So the second axis is the best model you can use inside the allowance.

![Scatter of usage per dollar against intelligence ceiling. Claude Ultra and Claude Team Premium sit at the top right at index 64.6. Claude Pro sits far right but lower, at 59.1 (Opus 5.5). ChatGPT plans sit at 62.0 between 10× and 16×. Copilot Pro+ and Max reach 64.6 but sit near 2×. Grey curves join plans with equal frontier value.](figures/fig4-intelligence-x-access.png "Usage per dollar against the best included model. Scope: same plans and measurements as the usage chart; Artificial Analysis Intelligence Index v5.0.1; as of 19 Nov 2026.")

*Sources: [Artificial Analysis](https://artificialanalysis.ai/models); [Wokey, Claude](https://wokey.ai/claude-usage-limits); [Wokey, Codex](https://wokey.ai/codex-usage-limits); [Inventive HQ](https://inventivehq.com/blog/codex-subscription-usage-measured); [ContextBolt](https://contextbolt.com/blog/codex-usage-limits/); [SemiAnalysis via explainx](https://www.explainx.ai/blog/semianalysis-claude-vs-chatgpt-subscription-api-value-limit-testing-2026); [GitHub Docs](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-individuals). Data: [CSV](https://careinstitute.ai/blog/ai-subscription-value-per-dollar/data/subscription-value-hard-numbers-2026-11-19.csv).*

Adding the model changes the top of the ranking:

- Claude Pro has the highest ratio, but its best model is Opus 5.5 (59.1). On Pro, Fable 5.2 is available only through paid extra usage.
- Claude Ultra has almost the same ratio *and* reaches Fable 5.2 (64.6). Its frontier value score is the highest among measured plans: **34.1**.
- ChatGPT plans reach GPT-6.2 Astra, OpenAI's best model on these plans (62.0, [AA](https://artificialanalysis.ai/models/gpt-6-2-astra)), on every tier from Plus up.
- Copilot Pro+ and Max also reach Fable 5.2, but their usage is worth only about 2× the price.

## How much total usage does each plan buy?

A ratio says nothing about scale. A $20 plan and a $200 plan can both return about 34×, yet one gives you about $700 of work a month and the other about $6,800. In the final chart, bubble area shows that monthly total. Half-shaded Claude bubbles show the Fable 5.2 limit: up to 50% of the weekly allowance can run on Fable, and the rest uses Opus 5.5.

![Bubble chart of usage per dollar against intelligence ceiling, with bubble area showing monthly API-priced usage. Claude Ultra is the largest bubble at the top right ($6,830 per month), half-shaded for the Fable 5.2 limit. Claude Team Premium ($3,108) and ChatGPT Pro $200 ($3,191) are similar in size. Claude Pro ($706) and ChatGPT Plus ($202) are small. Copilot bubbles are small and sit near 2×.](figures/fig5-full-picture.png "Frontier value, with bubble area showing monthly API-priced usage. Scope: full monthly use, agentic coding with heavy caching, API list prices. ChatGPT Pro $100 and $500 omitted (last measured before the Oct 29 increase). Google: no measured allowance. As of 19 Nov 2026.")

*Sources: [Artificial Analysis](https://artificialanalysis.ai/models); [Wokey, Claude](https://wokey.ai/claude-usage-limits); [Wokey, Codex](https://wokey.ai/codex-usage-limits); [Inventive HQ](https://inventivehq.com/blog/codex-subscription-usage-measured); [SemiAnalysis via explainx](https://www.explainx.ai/blog/semianalysis-claude-vs-chatgpt-subscription-api-value-limit-testing-2026); [GitHub Docs](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-individuals); [Anthropic Help Center](https://support.claude.com/en/articles/15424964-claude-fable-models-on-your-plan). Data: [CSV](https://careinstitute.ai/blog/ai-subscription-value-per-dollar/data/subscription-value-hard-numbers-2026-11-19.csv).*

| Plan | Frontier value score | Best model in allowance | Monthly usage |
|---|---:|---|---:|
| Claude Ultra $200 | **34.1** | Fable 5.2 (up to 50% of weekly limit), then Opus 5.5 | $6,830 |
| Claude Pro $20 | 32.3 | Opus 5.5 | $706 |
| Claude Team Premium $125 | 24.9 | Fable 5.2 (up to 50%), then Opus 5.5 | $3,108 |
| ChatGPT Pro $200 | 15.3 | GPT-6.2 Astra | $3,191 |
| ChatGPT Pro $100* | 13.9 | GPT-6.2 Astra | $1,450 |
| ChatGPT Pro $500* | 13.4 | GPT-6.2 Astra | $6,955 |
| ChatGPT Plus $20 | 9.7 | GPT-6.2 Astra | $202 |
| Copilot Max $100 | 2.0 | Fable 5.2 | $200 |
| Copilot Pro+ $39 | 1.8 | Fable 5.2 | $70 |
| Copilot Pro $10 | 1.3 | Sonnet 5.5 class | $15 |
{caption="Frontier value score by plan"}

*\*Measured before OpenAI's reported 29 October limit increase for Pro, so likely understated. Scores use the formula above and the sources in the usage table.*

## What is the best AI subscription at each price point?

Based on measured data only, as of 19 November 2026:

| Budget | Best measured choice | Why | Runner-up |
|---|---|---|---|
| About $20 | **Claude Pro** | 35× usage, Opus 5.5 | ChatGPT Plus: 10×, but a higher ceiling (GPT-6.2 Astra) |
| About $100–125 | **Claude Team Premium seat** ($125, team of 2+) | 25× with Fable 5.2 | ChatGPT Pro $100 (14×, measured before the increase) |
| About $200 | **Claude Ultra** | 34× with Fable 5.2, about $6,830/month | ChatGPT Pro $200: 16×, about $3,191/month |
| One plan from each lab, low cost | Claude Pro + ChatGPT Plus + Google AI Pro, about $60 | Three labs; $908/month of measured usage, plus Google's unmeasured share | — |
| One plan from each lab, heavy use | Claude Ultra + ChatGPT Pro $200 + Google AI Pro, about $420 | About $10,000/month of measured usage, plus Google | — |
{caption="Best measured AI subscription at each budget"}

Claude Max ($100) is missing from this table because we found no current independent measurement of it. The next section gives the figures we do have.

## What we don't know yet

We found no current independent measurement for the plans below, so **none of them is ranked above**. The table shows what is known about each.

| Plan | Price | What is known | Best available figure | Status |
|---|---|---|---|---|
| Claude Max | $100 | Anthropic says Max gives "5x" Pro usage per 5 hours and per week ([Help Center](https://support.claude.com/en/articles/11049741-what-is-the-max-plan)) | 5 × measured Pro = about $3,530 (35×). An October test measured $5,725 on Opus 5.5 ([SemiAnalysis via explainx](https://www.explainx.ai/blog/semianalysis-claude-vs-chatgpt-subscription-api-value-limit-testing-2026)) | Derived; a current tracker is still collecting data |
| Google AI Pro / Ultra | $19.99 / $99.99 / $199.99 | Includes Gemini 4 Neon (61.4) and Claude Opus 5.5 in Antigravity, Google's coding tool ([Antigravity plans](https://antigravity.google/docs/plans)) | A rule of thumb of about 4× price ([llmperks](https://llmperks.com/subscriptions)); not measured | Unknown |
| ChatGPT Pro $100 and $500 today | $100 / $500 | Limits reportedly doubled on 29 Oct ([ContextBolt](https://contextbolt.com/blog/codex-usage-limits/)) | October figures above | Likely understated |
| Muse Code (Meta) | $5–200 | Meta publishes tier multipliers only | One October test of an older model | Not comparable yet |
| X Subscription (xAI, Grok 4.7) | $8–200 | Grok 4.7 scores 35.5 on the index | Community estimates span more than 10× | Unknown |
| Kimi, GLM Coding Plan, Kiro | $18–200 | Quotas or credit values not published in dollars | None | Unknown |
{caption="Plans without a current independent measurement"}

If you have measured any of these, we would like to compare notes.

## The case against these numbers

- They are ceilings. The ratios assume you use the full allowance every week. A light user may pay less on the API than on any subscription.
- Caching inflates the dollar figure. The measurements came from coding agents that reuse most of their input. Work with less reuse, such as one-off document analysis, will show a smaller API-dollar figure.
- API dollars are not tasks. A model that uses fewer tokens per task finishes more work from the same allowance, so a lower API-dollar figure can still mean more finished tasks. The cost-per-task table above is a first look at this.
- The score credits the best model to the whole allowance. Claude Ultra and Team Premium are scored at Fable 5.2's index, but at most half of the weekly limit can run on Fable 5.2, and the usage measurements were mostly Opus 5.5. Read their scores as an upper bound.
- The Intelligence Index is a broad average. A model that ranks lower overall may do better on your work, so test with your own tasks.
- Limits change often. OpenAI reportedly raised Pro limits on 29 October, and Anthropic changed its plan structure this autumn. OpenAI is reported to be pooling ChatGPT and Codex usage from December. Treat every figure here as dated.
- Terms of use still apply. Plans are licensed to one person. Sharing accounts, reselling access or routing a plan through unofficial proxies can breach the vendor's terms. Use each vendor's own apps and tools.
- Sensitive data needs its own check. A subscription's value says nothing about whether you may send it patient, client or classified data. Before you send any protected data, check the vendor's terms and whether an agreement such as a business associate agreement is available for that plan. See [The Privacy-Routing Problem](/blog/the-privacy-routing-problem/).

## Frequently asked questions

### Which AI subscription is the best value in November 2026?
Among plans with independent measurements, Claude Ultra ($200) has the highest frontier value score (34.1): about $6,830 of API-priced usage a month, with up to half of it on Fable 5.2. At $20, Claude Pro returns the most usage per dollar (35×), but its best model is Opus 5.5.

### Is ChatGPT Pro or Claude Ultra better value at $200?
On measured data, Claude Ultra buys about twice the API-priced usage of ChatGPT Pro $200 (about $6,830 vs $3,191 a month) and reaches a higher-scoring model (Fable 5.2 at 64.6 vs GPT-6.2 Astra at 62.0). ChatGPT Pro is still the way to get OpenAI's models at volume.

### Is ChatGPT Pro worth it?
On measured data, ChatGPT Pro $200 returns about 16× its price in API-priced usage (about $3,191 a month), against about 10× for ChatGPT Plus. It is the way to use OpenAI's models at volume. At the same $200, Claude Ultra measured about twice as much usage (about $6,830 a month). ChatGPT Pro $100 and $500 were measured before OpenAI's reported 29 October limit increase, so their figures are likely understated.

### Is GitHub Copilot good value for heavy AI use?
Not for volume. Copilot credits are worth $0.01 each and are spent at API list prices, so Pro ($10), Pro+ ($39) and Max ($100) return 1.5 to 2× their price. Copilot is a low-cost way to reach many models, including Fable 5.2 on Pro+ and Max, but not a way to buy a large monthly allowance.

### Is a subscription cheaper than the API?
For heavy agentic coding, yes: measured Claude and ChatGPT plans return 10 to 35× their price at API list rates. For light or occasional use, pay-as-you-go API access can cost less. Copilot's credit plans are close to API pricing (1.5 to 2×).

### How much does it cost to run Fable 5.2 at max effort?
About $26.54 per task on the Artificial Analysis evaluation at list API prices, against $9.08 at high effort. Max effort adds 2.7 index points (61.8 to 64.6) for 2.9× the cost.

### Does Claude Max ($100) give 5× the usage of Pro?
Anthropic says Max gives 5× Pro usage per 5-hour window (a usage limit that resets every five hours) and per week. If that holds, Max returns about $3,530 a month (35×). We found no current independent measurement to confirm it, so we do not rank it.

### How much usage does Google AI Pro include?
Google does not publish the size of the coding allowance in its AI plans, and we found no independent measurement as of 19 November 2026. AI Pro does include Gemini 4 Neon, which scores 61.4 on the Intelligence Index.

### What is the frontier value score?
It is CARE Institute's combined measure: (API-priced usage ÷ subscription price) × (intelligence index of the best included model ÷ 64.6). A score of 30 means each $1 buys about $30 of usage at the level of the top-scoring model.

## Sources

**Intelligence and cost per task:** Artificial Analysis Intelligence Index v5.0.1 model pages, accessed 19 Nov 2026: [Fable 5.2](https://artificialanalysis.ai/models/claude-fable-5-2), [Opus 5.5](https://artificialanalysis.ai/models/claude-opus-5-5), [GPT-6.2 Astra](https://artificialanalysis.ai/models/gpt-6-2-astra), [GPT-6 Astra](https://artificialanalysis.ai/models/gpt-6-astra), [GPT-6.1 Sol](https://artificialanalysis.ai/models/gpt-6-1-sol), [Gemini 4 Neon](https://artificialanalysis.ai/models/gemini-4-neon), [Gemini 4 Carbon](https://artificialanalysis.ai/models/gemini-4-carbon), [Grok 4.7](https://artificialanalysis.ai/models/grok-4-7).

**Subscription usage measurements:** [Wokey, Claude usage limits](https://wokey.ai/claude-usage-limits) (9 to 16 Nov 2026); [Wokey, Codex usage limits](https://wokey.ai/codex-usage-limits) (9 to 16 Nov 2026); [SemiAnalysis, Anthropic Subscriptions Offer 5x+ More Value Than OpenAI](https://newsletter.semianalysis.com/p/anthropic-subscriptions-offer-5x) (5 Oct 2026, partly paywalled) and its open [explainx summary](https://www.explainx.ai/blog/semianalysis-claude-vs-chatgpt-subscription-api-value-limit-testing-2026); [Inventive HQ, Codex subscription usage measured](https://inventivehq.com/blog/codex-subscription-usage-measured) (4 Oct 2026); [ContextBolt, Codex usage limits](https://contextbolt.com/blog/codex-usage-limits/).

**Vendor plan pages:** [Claude pricing](https://claude.com/pricing); [What is the Max plan](https://support.claude.com/en/articles/11049741-what-is-the-max-plan); [What is the Ultra plan](https://support.claude.com/en/articles/17728952-what-is-the-ultra-plan); [Claude Fable models on your plan](https://support.claude.com/en/articles/15424964-claude-fable-models-on-your-plan); [What is the Team plan](https://support.claude.com/en/articles/9266767-what-is-the-team-plan); [Codex pricing](https://learn.chatgpt.com/docs/pricing); [GitHub Copilot plans](https://docs.github.com/en/copilot/about-github-copilot/plans-for-github-copilot); [Copilot usage-based billing](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-individuals); [Google AI plans](https://one.google.com/about/google-ai-plans/); [Antigravity plans](https://antigravity.google/docs/plans).

**Data:** [subscription-value-hard-numbers-2026-11-19.csv](https://careinstitute.ai/blog/ai-subscription-value-per-dollar/data/subscription-value-hard-numbers-2026-11-19.csv) · [effort-curves-2026-11-19.csv](https://careinstitute.ai/blog/ai-subscription-value-per-dollar/data/effort-curves-2026-11-19.csv). Each row names its measurement and source.

*Prices are US month-to-month list prices as of 19 November 2026. Annual discounts are excluded.*
