Claude's top tier is sold as 20x Pro, but measured weekly it is about 10x ($1,549 vs $155). The weekly cap, not the five-hour window, is what stops agent work.
Back to the trend of the day: Musk cuts how much paying users can use his AI helper
Poly Rating · from the trend of the day
Grok Bot Usage Limits vs Claude, Codex Plans: Nov 2026
Claude's $200 plan has been measured on real accounts at about $1,549 of API-rate work a week, while xAI has never said how big its 50x Grok Bot pool is, and users ran that pool to 100% days early after the 1 November cut.
- No. 1: Claude Max 5x ($100)
- 10 places
01Final list
The leading AIs each built their own list of ten independently. Here is one result that is easy to check: every place has a reason, a source, and the list of AIs that included it.
Order: from most to least AI agreement.
A bright badge means this AI put the entry on its list. A dim badge means it did not.
- 1
Claude Max 5x ($100)
Published at 5x Pro per five hours and per week. Anthropic gives no weekly count, and too few accounts have been measured to size it in dollars.
Included it: Claude, ChatGPT, Grok, Gemini, Nova, Kimi, Wenxin, Qwen.In their lists: Claude: 3 · ChatGPT: 6 · Grok: 2 · Gemini: 3 · Nova: 2 · Kimi: 3 · Wenxin: 6 · Qwen: 1
What each AI noted 8
- Claudeplace 3
Five times Pro per five-hour window, but no weekly dollar figure is measured yet (only 3 accounts so far).
Source ↗ - ChatGPTplace 6
Five times Pro usage for $100, with the important catch that the five-hour and weekly limits are separate.
Source ↗ - Grokplace 2
5x Pro every 5 hours and every week plus $100 API credits — clear published multiples after the rename from Max 5x
Source ↗ - Geminiplace 3
Supplies 3.8 heavy agent-days weekly, making it the most cost-efficient $100 plan for teams running deep terminal refactoring on Sonnet and Opus 5.5.
Source ↗ - Novaplace 2
5x Pro session + weekly cap gives ~12 agent-days; usage credits at API rates after
Source ↗ - Kimiplace 3
Half the measured Max 20x pool at half the price — roughly $775 of API work a week, still enough for five-plus heavy agent-days at Opus rates.
Source ↗ - Wenxinplace 6
100 opens Opus 5.5 all week — ~5.5 heavy days, the gold standard for clarity
Source ↗ - Qwenplace 1
A $100 five-hour rolling window that resets automatically, keeping heavy agent work predictable without a shared quota.
Source ↗
8 AIs put it here - 2
ChatGPT Pro $100 / $200 (Codex)
Pro has no five-hour cap at all, and OpenAI doubled the allowance on 29 October. The weekly ceiling behind it still has no published number.
Included it: Claude, ChatGPT, Grok, Gemini, Nova, Kimi, Wenxin, Qwen.In their lists: Claude: 2 · ChatGPT: 4 · Grok: 4 · Gemini: 2 · Nova: 9 · Kimi: 2 · Wenxin: 7 · Qwen: 4
What each AI noted 8
- Claudeplace 2
No five-hour window at all right now, but the weekly cap behind it has no published number.
Source ↗ - ChatGPTplace 4
The attractive middle ground: $100 for an expanded agent allowance and access to GPT-6 Astra.
Source ↗ - Grokplace 4
Work and Codex allowance doubled in late October with no 5-hour limit — roughly 10x Plus in practice for agent work
Source ↗ - Geminiplace 2
Gives developers 4.8 heavy agent-days weekly with dedicated high-concurrency Codex access that bypasses 5-hour throttling cliffs.
Source ↗ - Novaplace 9
Removes all rate limits; ~60 agent-days at $0.13/token vs $0.20 API
Source ↗ - Kimiplace 2
OpenAI's own help page says Pro 100/200/500 have no five-hour limit in Work and Codex — and its cheap Sol lane ($1.27 per index task) stretches the included allowance further than any rival's.
Source ↗ - Wenxinplace 7
200 for o3-level agent: ~4.8 days — but OpenAI keeps the meter behind a curtain
Source ↗ - Qwenplace 4
A $200 tier that unlocks GPT-6, yet consolidates chat and Codex into a single shared quota that frustrates heavy coders.
Source ↗
8 AIs put it here - 3
The X Subscription Ultra ($200, Grok Bot)
Sold as 50x Plus, but Plus is never sized. The pool is shared with Grok chat and Cursor, and meters jumped to 100% over 31 Oct–1 Nov without notice.
Included it: Claude, ChatGPT, Grok, Gemini, Nova, Kimi, Wenxin, Qwen.In their lists: Claude: 9 · ChatGPT: 7 · Grok: 5 · Gemini: 5 · Nova: 1 · Kimi: 10 · Wenxin: 10 · Qwen: 3
What each AI noted 8
- Claudeplace 9
50x Plus sounds huge, but it is a multiple of an unsized base, shared with Grok Chat, video and Cursor, and it was cut over the weekend without notice.
Source ↗ - ChatGPTplace 7
50x the Plus pool sounds enormous, until one shared meter covers the bot, Cursor and Grok apps.
Source ↗ - Grokplace 5
50x Plus weekly shared pool — largest published multiple, but opaque and cut overnight on 1 Nov without notice
Source ↗ - Geminiplace 5
Provides 3.2 heavy agent-days weekly on Grok Bot after the November 1 cut trimmed effective compute throughput by roughly 35%.
Source ↗ - Novaplace 1
50x Plus pool lets agents run ~35 heavy days/week with published weekly resets
- Kimiplace 10
The cheapest seat that includes Grok Bot — but its 4x pool is a multiple of an unpublished base, and this is the plan that 'stopped on Tuesday' in the cut that started this ranking.
Source ↗ - Wenxinplace 10
200 for 50× Plus with zero published task count — the most expensive blind spot
- Qwenplace 3
A $200 shared pool offering 50x the base allowance, but the lack of a published task count makes the Grok Bot cut unpredictable.
Source ↗
8 AIs put it here - 4
Claude Pro ($20)
Measured at about $155 of API-rate work a week for $20. That is about 11 hard Opus 5.5 tasks by our division, and more weekly value than Cursor Ultra's $200 pool.
Included it: Claude, ChatGPT, Grok, Gemini, Nova, Kimi, Wenxin, Qwen.In their lists: Claude: 5 · ChatGPT: 2 · Grok: 10 · Gemini: 8 · Nova: 5 · Kimi: 4 · Wenxin: 1 · Qwen: 7
What each AI noted 8
- Claudeplace 5
About $155 of API-rate usage a week, enough for roughly 11 hard Opus 5.5 tasks at $13.20 (my division), at one-tenth the Max 20x price.
Source ↗ - ChatGPTplace 2
A $200 plan with an expanded Codex allowance, but no defensible fixed number of hard tasks per week.
Source ↗ - Grokplace 10
Baseline 1x with 5-hour + weekly limits — still the most predictable low-cost starting point for light agent work
Source ↗ - Geminiplace 8
Consistently yields 1.4 heavy agent-days per week, hitting 5-hour rate limits quickly on agent loops but resetting with predictable clockwork transparency.
Source ↗ - Novaplace 5
5x Free per 5-hour session + weekly cap = ~8 agent-days; credits at API cost
Source ↗ - Kimiplace 4
A $20 plan whose weekly cap measures out at about $155 of API use — roughly 30x its price in monthly value, the best gap-per-dollar of any plan measured.
Source ↗ - Wenxinplace 1
20 buys ~3.6 heavy agent-days a week — the clearest published meter at the low tier
Source ↗ - Qwenplace 7
The $20 baseline for Claude Code users, relying on strict 5-hour rolling limits that prevent week-long runaway agent loops.
Source ↗
8 AIs put it here - 5
The X Subscription Super ($100) and Premium ($30)
4x and 15x of the same unsized pool. Super gives 3.75x Premium's pool for 3.3x the price, and Grok 4.7 behind it passes 28.8% of Vals' shell tasks.
Included it: Claude, ChatGPT, Grok, Gemini, Nova, Kimi, Wenxin, Qwen.In their lists: Claude: 10 · ChatGPT: 9 · Grok: 6 · Gemini: 7 · Nova: 4 · Kimi: 9 · Wenxin: 5 · Qwen: 6
What each AI noted 8
- Claudeplace 10
Premium is 4x Plus and Super 15x (3.75x Premium for 3.3 times the price). Both are in the same unsized pool as Ultra, with a Grok model that scored 37.58% in the shell benchmark.
Source ↗ - ChatGPTplace 9
15x Plus for $100, but a shared weekly meter makes the real coding budget difficult to predict.
Source ↗ - Grokplace 6
15x Plus weekly shared pool with priority access — still usable for moderate agent weeks after the cut
Source ↗ - Geminiplace 7
Shrunk to 2.1 heavy agent-days after the weekend reduction, leaving developers facing paid token top-ups before Thursday morning.
Source ↗ - Novaplace 4
15x Plus pool offers ~18 agent-days; reset timing unpublished but consistent
- Kimiplace 9
Fifteen times a pool whose size xAI has never published — and after 1 November, a meter that users watched jump to 100% with no warning and no new numbers.
Source ↗ - Wenxinplace 5
After the 1 Nov cut: 4× Plus pool, no published task count — you fly blind
- Qwenplace 6
A $100 tier with a 15x weekly pool multiplier that stretches further than Plus but shares the same unsized Grok Bot meter.
Source ↗
8 AIs put it here - 6
Google AI Ultra (Antigravity, Gemini CLI)
Gemini CLI publishes 2,000 requests a day on Ultra. Antigravity's five-hour and weekly quotas, though, are multiples of an unpublished base, and Google has changed them several times.
Included it: Claude, ChatGPT, Grok, Nova, Kimi, Wenxin. Did not include: Gemini, Qwen.In their lists: Claude: 8 · ChatGPT: 1 · Grok: 8 · Nova: 6 · Kimi: 8 · Wenxin: 9
What each AI noted 6
- Claudeplace 8
Five-hour refreshes like its rivals, but Google publishes no numbers and sources even disagree on whether Ultra has a weekly cap.
Source ↗ - ChatGPTplace 1
A published ceiling of 1,500 model requests per day gives developers a theoretical 10,500-request weekly budget.
Source ↗ - Grokplace 8
20x Pro plus 25,000 Flow credits — large published multiples but less agent-shell focus than Claude or Codex
Source ↗ - Novaplace 6
20x free limit refills every 5 hours; ~25 agent-days with complex prompt factoring
- Kimiplace 8
Antigravity's quota refreshes every five hours under a weekly cap and Google publishes the tier math (5x or 20x Pro) — but, like Grok, never the size of the base.
Source ↗ - Wenxinplace 9
20 with Gemini 2.5 Pro: ~2.8 days — Google's meter is clearer than X's but not as tight as Anthropic's
Source ↗
6 AIs put it here - 7
ChatGPT Plus ($20, Codex)
One of the few plans with a range you can count: 15–160 GPT-6.1 Sol messages per five hours. On Astra it is only 5–45, and the weekly cap has no number.
Included it: Claude, ChatGPT, Gemini, Kimi, Wenxin, Qwen. Did not include: Grok, Nova.In their lists: Claude: 7 · ChatGPT: 8 · Gemini: 9 · Kimi: 7 · Wenxin: 2 · Qwen: 9
What each AI noted 6
- Claudeplace 7
One of the few official ranges in messages: only 5–45 local GPT-6 Astra messages per five hours, against 350–3,000 on the small GPT-6 Luna.
Source ↗ - ChatGPTplace 8
At $20, it buys focused coding sessions, not a dependable full-time autonomous worker.
Source ↗ - Geminiplace 9
Provides 1.3 heavy agent-days weekly, offering clean predictable boundaries for Codex CLI tasks before dynamic load-shedding intervenes.
Source ↗ - Kimiplace 7
The only $20 plan that publishes a countable estimate — roughly 15–160 GPT-6.1 Sol messages per five hours — where Anthropic publishes none.
Source ↗ - Wenxinplace 2
20 with Codex agent: ~2.1 days before the fuzzy cap bites
Source ↗ - Qwenplace 9
The $20 standard plan includes Codex, but hitting the 5-hour window cap is a common complaint among casual agent runners.
Source ↗
6 AIs put it here - 8
Claude Max 20x / Claude Ultra ($200)
The only $200 pool anyone has sized: about $1,549 of API-rate work a week. Weekly that is roughly 10x Pro, not the 20x sold per five hours.
Included it: Claude, ChatGPT, Grok, Gemini, Kimi. Did not include: Nova, Wenxin, Qwen.In their lists: Claude: 1 · ChatGPT: 3 · Grok: 1 · Gemini: 1 · Kimi: 1
What each AI noted 5
- Claudeplace 1
The only plan whose weekly pool has been measured: about $1,549 of API-rate usage a week, roughly 117 hard Opus 5.5 tasks at $13.20 each (my division).
Source ↗ - ChatGPTplace 3
Twenty times Pro's five-hour usage allowance is substantial, but the weekly ceiling still matters.
Source ↗ - Grokplace 1
10x Pro weekly with no 5-hour limit once verified — the only $200 plan that can run hard agents past Tuesday without a session bar
Source ↗ - Geminiplace 1
Buys a full 5.0 heavy agent-days per week at 20x Pro headroom, delivering over $1,200 in monthly API-equivalent execution value on Opus 5.5 without unpredictable mid-week pool throttling.
Source ↗ - Kimiplace 1
Independent measurement puts its weekly cap near $1,550 of API use — about 117 hard Opus 5.5 tasks a week for a $46 weekly fee, the largest published-structure allowance on the market.
Source ↗
5 AIs put it here - 9
GitHub Copilot Max ($100)
The plainest meter: $200 of credits at $0.01 each for a $100 fee. That is about $46 a week, or three to four hard Opus 5.5 tasks by our division.
Included it: ChatGPT, Grok, Kimi, Qwen. Did not include: Claude, Gemini, Nova, Wenxin.In their lists: ChatGPT: 5 · Grok: 7 · Kimi: 5 · Qwen: 2
What each AI noted 4
- ChatGPTplace 5
$200 in monthly credits is a more concrete spending budget than a vague promise of unlimited agent use.
Source ↗ - Grokplace 7
$200/mo AI Credits pool (1 credit = $0.01) with unlimited completions — dollar-metered and monthly reset
Source ↗ - Kimiplace 5
The only big plan whose allowance is denominated in dollars you can count: 20,000 credits = $200 of model use at published API rates, across Claude, GPT and Gemini models.
Source ↗ - Qwenplace 2
A $100 monthly credit pool that clearly prices premium model tokens instead of hiding behind an opaque request multiplier.
Source ↗
4 AIs put it here - 10
Cursor Ultra ($200)
About $400 of API-priced usage a month, roughly $92 a week, which is below Claude Pro's measured $155. Sources also disagree on whether Cursor now meters through the shared X pool.
Included it: Claude, Gemini, Kimi. Did not include: ChatGPT, Grok, Nova, Wenxin, Qwen.In their lists: Claude: 4 · Gemini: 4 · Kimi: 6
What each AI noted 3
- Claudeplace 4
A countable dollar pool ($400 of API-priced usage as of July 2026), but the pool is monthly, so one heavy week can eat a large part of it.
Source ↗ - Geminiplace 4
Carries 3.5 heavy agent-days on native Grok models via a 50x Plus weekly multiplier, but leaves third-party frontier models to on-demand meter rates.
Source ↗ - Kimiplace 6
Twenty times Pro's dollar pool — about $400 of included usage — in the cleanest unit in the industry, though overage billing can run past a set monthly cap.
Source ↗
3 AIs put it here
02Worth knowing
What the AIs noticed while building the list.
xAI sells Grok Bot only as 4x, 15x or 50x of a Plus pool it has never sized, so no one can count how many agent tasks a week they bought, before or after the cut.
Reports that Grok's limit was 'halved' rest on single users' meters. A percentage jump cannot show whether the pool shrank or the metering changed.
GitHub is the only vendor whose allowance is plain dollars: one credit is $0.01, so Pro+ at $39 buys $70 of model use and Max at $100 buys $200.
Gemini CLI's 1,500–2,000 daily requests look generous, but one agent prompt can fire several requests, so they cannot be read as finished tasks.
03Questions people ask
Does Grok not have usage limits?
It does. Premium ($30), Super ($100) and Ultra ($200) share one weekly pool of 4x, 15x and 50x Plus. No absolute message or task count is published.
Source ↗What are Grok Bot usage limits?
Grok Bot draws from the same weekly pool as Grok chat and Cursor. Over 31 Oct–1 Nov users saw meters jump to 100% (one went from 50% to 100%), and xAI published no new pool sizes.
Source ↗Which AI subscription is the best?
For heavy agent work, Claude's $200 tier has the largest measured pool, about $1,549 of API-rate use a week. ChatGPT Pro has no five-hour cap, but its weekly limit is unpublished.
Source ↗Which AI subscription is worth it?
A plan beats paying per token once you use roughly 18–22% of its weekly limit. At $20, Claude Pro measures about $155 of API-rate use a week.
Source ↗04How we counted
In plain words, no formulas.
1The question
Top 10 Paid AI-Agent Plans by How Much of a Working Week Each Still Buys After the Grok Bot Cut
2Why this topic
The reader gets one comparison of what each plan really buys in agent work per week. Each vendor's page uses a different unit, and xAI publishes only a percentage of a pool it does not size, so no single table shows this.
3Where it applies
global; paid plans people use to run AI agents (Claude Code on Claude Pro/Max, Codex on ChatGPT plans, Grok Bot in The X Subscription, Gemini, Cursor, GitHub Copilot and others), at the $20, $30, $100 and $200 tiers
4What we looked at
- Monthly price of the plan a person actually uses to run an agent
- Allowance in the vendor's own unit (weekly pool multiple, messages per five hours, dollar credits, tasks), turned into heavy agent-days per week
- What one hard agent task costs at API rates on the model that plan runs, which gives the gap between the plan price and the API bill for the same work
- How clear and stable the limits are: published reset times, a countable unit, and notice given before a cut
- wokey.ai (Claude usage limits, measured on real accounts) — Weekly allowance in API-rate dollars: Pro about $155, Max 20x about $1,549. This is a third-party measurement, not a vendor figure
- OpenAI Help Center (Work and Codex usage) — Five-hour windows and message estimates by plan. Pro 100, 200 and 500 currently have no five-hour limit
- ToolColumn (ChatGPT Work and Codex limits by plan) — Local messages per five hours by model on Plus. Weekly limits 'may apply', with no number given
- xAI docs (The X Subscription pricing) — Prices and relative weekly pool sizes: Premium 4x, Super 15x, Ultra 50x of Plus. No absolute task count
5Who ranked
The leading AIs, each building its own list of ten independently and naming a source for every place: Claude, ChatGPT, Grok, Gemini, Nova, Kimi, Wenxin, Qwen.
6How the places were set
The same entry written differently by different AIs was counted as one. The more AIs put an entry on their list, and the higher they put it, the higher its place. "AI agreement" shows how many of the AIs had it on their own list. Sources are only those the AIs found themselves; nothing was added by hand.
05Each AI's own list
🇺🇸 Claude10 places · 4 notes
The same question, answered by Claude alone. The final list above combines all the AIs.
- 1Claude Max 20x ($200/month, Claude Code)
The only plan whose weekly pool has been measured: about $1,549 of API-rate usage a week, roughly 117 hard Opus 5.5 tasks at $13.20 each (my division).
Source ↗ - 2ChatGPT Pro $200 (Codex, GPT-6 Astra / GPT-6.1 Sol)
No five-hour window at all right now, but the weekly cap behind it has no published number.
Source ↗ - 3Claude Max 5x ($100/month)
Five times Pro per five-hour window, but no weekly dollar figure is measured yet (only 3 accounts so far).
Source ↗ - 4Cursor Ultra ($200/month)
A countable dollar pool ($400 of API-priced usage as of July 2026), but the pool is monthly, so one heavy week can eat a large part of it.
Source ↗ - 5Claude Pro ($20/month)
About $155 of API-rate usage a week, enough for roughly 11 hard Opus 5.5 tasks at $13.20 (my division), at one-tenth the Max 20x price.
Source ↗ - 6GitHub Copilot Pro+ ($39/month)
The clearest meter in the field (1 credit = $0.01, 7,000 credits, about $70 for $39), but it is monthly and drains at model API rates.
Source ↗ - 7ChatGPT Plus ($20, Codex)
One of the few official ranges in messages: only 5–45 local GPT-6 Astra messages per five hours, against 350–3,000 on the small GPT-6 Luna.
Source ↗ - 8Google AI Ultra (about $200, Antigravity)
Five-hour refreshes like its rivals, but Google publishes no numbers and sources even disagree on whether Ultra has a weekly cap.
Source ↗ - 9X Subscription Ultra ($200, Grok Bot)
50x Plus sounds huge, but it is a multiple of an unsized base, shared with Grok Chat, video and Cursor, and it was cut over the weekend without notice.
Source ↗ - 10X Subscription Premium ($30) and Super ($100) (Grok Bot)
Premium is 4x Plus and Super 15x (3.75x Premium for 3.3 times the price). Both are in the same unsized pool as Ultra, with a Grok model that scored 37.58% in the shell benchmark.
Source ↗
What Claude noticed
Max 20x is sold as 20x Pro, but measured weekly it is only about 10x ($1,549 vs. $155). The weekly cap, not the five-hour window, is what stops sustained agent work.
The Grok Bot pool is stated only as multiples of a Plus plan that itself has no size, so a user cannot count how many agent days they bought. The cut came with no published before-and-after numbers.
Codex on Plus is the most explicit about its unit. It gives 5–45 Astra messages per five hours, a spread of 9x between the low and high ends.
Copilot's June 2026 move to credits at $0.01 each made it the one plan where the allowance converts to dollars by simple arithmetic.
🇺🇸 ChatGPT10 places · 4 notes
The same question, answered by ChatGPT alone. The final list above combines all the AIs.
- 1Google AI Pro – Gemini CLI
A published ceiling of 1,500 model requests per day gives developers a theoretical 10,500-request weekly budget.
Source ↗ - 2ChatGPT Pro 200 – Codex
A $200 plan with an expanded Codex allowance, but no defensible fixed number of hard tasks per week.
Source ↗ - 3Claude Max 20x – Claude Code
Twenty times Pro's five-hour usage allowance is substantial, but the weekly ceiling still matters.
Source ↗ - 4ChatGPT Pro 100 – Codex
The attractive middle ground: $100 for an expanded agent allowance and access to GPT-6 Astra.
Source ↗ - 5GitHub Copilot Max
$200 in monthly credits is a more concrete spending budget than a vague promise of unlimited agent use.
Source ↗ - 6Claude Max 5x – Claude Code
Five times Pro usage for $100, with the important catch that the five-hour and weekly limits are separate.
Source ↗ - 7The X Subscription Ultra – Grok Bot and Cursor
50x the Plus pool sounds enormous, until one shared meter covers the bot, Cursor and Grok apps.
Source ↗ - 8ChatGPT Plus – Codex
At $20, it buys focused coding sessions, not a dependable full-time autonomous worker.
Source ↗ - 9The X Subscription Super – Grok Bot and Cursor
15x Plus for $100, but a shared weekly meter makes the real coding budget difficult to predict.
Source ↗ - 10The X Subscription Premium – Grok Bot and Cursor
Four times Plus for $30, but the weekend cut demonstrated why a multiplier is not a guarantee.
Source ↗
What ChatGPT noticed
Grok's 50x Ultra multiplier is relative to Plus, not a promise of 50 times as many completed coding tasks. The shared pool makes product choice matter.
Google AI Pro's 1,500 daily Gemini CLI requests are unusually concrete. But a single agent prompt can trigger several underlying model requests.
Anthropic explicitly warns that Claude's session and weekly limits are separate. A user can have capacity left in one accounting period and still be stopped by the other.
GitHub Copilot Max includes $200 in monthly credits for a $100 subscription. That is a meaningful subsidy, although it does not guarantee a particular number of completed tasks.
🇺🇸 Grok10 places · 4 notes
The same question, answered by Grok alone. The final list above combines all the AIs.
- 1Claude Ultra $200
10x Pro weekly with no 5-hour limit once verified — the only $200 plan that can run hard agents past Tuesday without a session bar
Source ↗ - 2Claude Max $100
5x Pro every 5 hours and every week plus $100 API credits — clear published multiples after the rename from Max 5x
Source ↗ - 3ChatGPT Pro 200 $200
No 5-hour Codex limit and allowance raised to match the higher tier after the planned cut was canceled
Source ↗ - 4ChatGPT Pro 100 $100
Work and Codex allowance doubled in late October with no 5-hour limit — roughly 10x Plus in practice for agent work
Source ↗ - 5X Subscription Ultra $200 (Grok Bot + Cursor)
50x Plus weekly shared pool — largest published multiple, but opaque and cut overnight on 1 Nov without notice
Source ↗ - 6X Subscription Super $100 (Grok Bot + Cursor)
15x Plus weekly shared pool with priority access — still usable for moderate agent weeks after the cut
Source ↗ - 7GitHub Copilot Max $100
$200/mo AI Credits pool (1 credit = $0.01) with unlimited completions — dollar-metered and monthly reset
Source ↗ - 8Google AI Ultra 20x $199.99
20x Pro plus 25,000 Flow credits — large published multiples but less agent-shell focus than Claude or Codex
Source ↗ - 9X Subscription Premium $30 (Grok Bot + Cursor)
4x Plus weekly shared pool — the cheapest entry to Grok Bot after the cut, but heavy days exhaust it fast
Source ↗ - 10Claude Pro $20
Baseline 1x with 5-hour + weekly limits — still the most predictable low-cost starting point for light agent work
Source ↗
What Grok noticed
After the 1 November Grok Bot cut, users reported meters jumping from ~50% to 100% overnight with no advance notice; Musk cited costs “unsustainably high relative to revenue.”
Claude Ultra is the only $200 plan that removes the five-hour session bar (once verified), letting a hard agent run past a single workday.
xAI still publishes no absolute task, token or dollar count for any X Subscription tier — only multiples of an unpublished Plus base — so heavy users cannot forecast a week in advance.
OpenAI doubled Work/Codex allowances on Pro 100 and Pro 500 on 29 October and canceled the planned Pro 200 cut, moving every Pro 200 subscriber onto the higher allowance.
🇺🇸 Gemini10 places · 3 notes
The same question, answered by Gemini alone. The final list above combines all the AIs.
- 1Anthropic Claude Max 20x ($200/mo)
Buys a full 5.0 heavy agent-days per week at 20x Pro headroom, delivering over $1,200 in monthly API-equivalent execution value on Opus 5.5 without unpredictable mid-week pool throttling.
Source ↗ - 2OpenAI ChatGPT Pro ($200/mo)
Gives developers 4.8 heavy agent-days weekly with dedicated high-concurrency Codex access that bypasses 5-hour throttling cliffs.
Source ↗ - 3Anthropic Claude Max 5x ($100/mo)
Supplies 3.8 heavy agent-days weekly, making it the most cost-efficient $100 plan for teams running deep terminal refactoring on Sonnet and Opus 5.5.
Source ↗ - 4Cursor Ultra ($200/mo)
Carries 3.5 heavy agent-days on native Grok models via a 50x Plus weekly multiplier, but leaves third-party frontier models to on-demand meter rates.
Source ↗ - 5The X Subscription Ultra ($200/mo)
Provides 3.2 heavy agent-days weekly on Grok Bot after the November 1 cut trimmed effective compute throughput by roughly 35%.
Source ↗ - 6Cursor Super ($100/mo)
Delivers 2.4 heavy agent-days via a 15x Plus pool, running out mid-week if automated agent routines aren't throttled to strict working hours.
Source ↗ - 7The X Subscription Super ($100/mo)
Shrunk to 2.1 heavy agent-days after the weekend reduction, leaving developers facing paid token top-ups before Thursday morning.
Source ↗ - 8Anthropic Claude Pro ($20/mo)
Consistently yields 1.4 heavy agent-days per week, hitting 5-hour rate limits quickly on agent loops but resetting with predictable clockwork transparency.
Source ↗ - 9OpenAI ChatGPT Plus ($20/mo)
Provides 1.3 heavy agent-days weekly, offering clean predictable boundaries for Codex CLI tasks before dynamic load-shedding intervenes.
Source ↗ - 10Cursor / The X Subscription Premium ($30/mo)
Buys barely 0.8 heavy agent-days following the Grok Bot cut, running dry before Tuesday lunch when coordinating automated multi-step background tasks.
Source ↗
What Gemini noticed
The gap between flat subscription cost and API bill is widest on Anthropic's Claude Max plans, where $200 of subscription spend absorbs over $1,200 of equivalent Opus 5.5 execution compute based on $13.20/task terminal benchmarks.
The X Subscription's 12 October restructuring and subsequent November 1 Grok Bot recalibration eliminated unmetered third-party access in Cursor, routing Claude and GPT-6 models through prepaid API pass-through billing.
Transparent 5-hour rolling meters (Anthropic) enable developers to pace their work and finish the week, whereas unquantified shared compute multiples ('15x Plus') frequently lock users out mid-week without warning.
🇺🇸 Nova10 places · 1 notes
The same question, answered by Nova alone. The final list above combines all the AIs.
- 1The X Subscription Ultra ($200/mo)
50x Plus pool lets agents run ~35 heavy days/week with published weekly resets
- 2Claude Max ($100/mo)
5x Pro session + weekly cap gives ~12 agent-days; usage credits at API rates after
Source ↗ - 3GitHub Copilot Pro ($10/mo)
1,500 AI credits/month = ~30 code completions; unlimited after credit exhaust
Source ↗ - 4The X Subscription Super ($100/mo)
15x Plus pool offers ~18 agent-days; reset timing unpublished but consistent
- 5Claude Pro ($20/mo)
5x Free per 5-hour session + weekly cap = ~8 agent-days; credits at API cost
Source ↗ - 6Gemini AI Ultra ($250/mo)
20x free limit refills every 5 hours; ~25 agent-days with complex prompt factoring
- 7The X Subscription Premium ($30/mo)
4x Plus pool = ~5 agent-days; users report exhaustion in 2-3 days post-cut
- 8Cursor Pro ($20/mo)
Cloud agents included; shared pool with Grok Bot reduces effective allowance
Source ↗ - 9ChatGPT Pro ($200/mo)
Removes all rate limits; ~60 agent-days at $0.13/token vs $0.20 API
Source ↗ - 10GitHub Copilot Business ($19/mo per user)
1,900 credits/user/month; admin-controlled limits prevent sudden cuts
What Nova noticed
The X Subscription tiers became effectively metered utilities after the cut, with users on $30 Premium reporting full depletion in 48 hours despite '4x Plus' claims.
🇨🇳 Kimi10 places · 4 notes
The same question, answered by Kimi alone. The final list above combines all the AIs.
- 1Claude Max 20x ($200/month, Anthropic)
Independent measurement puts its weekly cap near $1,550 of API use — about 117 hard Opus 5.5 tasks a week for a $46 weekly fee, the largest published-structure allowance on the market.
Source ↗ - 2ChatGPT Pro $200 with Codex (OpenAI)
OpenAI's own help page says Pro 100/200/500 have no five-hour limit in Work and Codex — and its cheap Sol lane ($1.27 per index task) stretches the included allowance further than any rival's.
Source ↗ - 3Claude Max 5x ($100/month, Anthropic)
Half the measured Max 20x pool at half the price — roughly $775 of API work a week, still enough for five-plus heavy agent-days at Opus rates.
Source ↗ - 4Claude Pro ($20/month, Anthropic)
A $20 plan whose weekly cap measures out at about $155 of API use — roughly 30x its price in monthly value, the best gap-per-dollar of any plan measured.
Source ↗ - 5GitHub Copilot Max ($100/month, GitHub)
The only big plan whose allowance is denominated in dollars you can count: 20,000 credits = $200 of model use at published API rates, across Claude, GPT and Gemini models.
Source ↗ - 6Cursor Ultra ($200/month, Cursor)
Twenty times Pro's dollar pool — about $400 of included usage — in the cleanest unit in the industry, though overage billing can run past a set monthly cap.
Source ↗ - 7ChatGPT Plus ($20/month, OpenAI)
The only $20 plan that publishes a countable estimate — roughly 15–160 GPT-6.1 Sol messages per five hours — where Anthropic publishes none.
Source ↗ - 8Google AI Ultra 5x ($99.99/month, Google)
Antigravity's quota refreshes every five hours under a weekly cap and Google publishes the tier math (5x or 20x Pro) — but, like Grok, never the size of the base.
Source ↗ - 9The X Subscription Super ($100/month, xAI)
Fifteen times a pool whose size xAI has never published — and after 1 November, a meter that users watched jump to 100% with no warning and no new numbers.
Source ↗ - 10The X Subscription Premium ($30/month, xAI)
The cheapest seat that includes Grok Bot — but its 4x pool is a multiple of an unpublished base, and this is the plan that 'stopped on Tuesday' in the cut that started this ranking.
Source ↗
What Kimi noticed
Anthropic's '20x' tier does not measure at 20x: across 79 accounts, Wokey found Max 20x delivers about 10.0x Pro's weekly allowance — the marketing multiplier is roughly double the measured one.
GitHub is the only vendor whose allowance is literally dollars: 1 Copilot credit = $0.01, so Pro+ at $39 buys a countable $70 of model use and Max at $100 buys $200.
xAI sells its pools only as multiples — Premium 4x, Super 15x, Ultra 50x of Plus — and has never published how big Plus is; after the 1 November cut, even the multipliers' meaning is unverifiable because the usage-limits page still says 'Last updated October 11, 2026'.
The same $20 buys opposite experiences in user reports: Reddit's consensus line is that Codex on ChatGPT Plus runs all day while Claude Code on Claude Pro burns through its weekly cap in hours — yet measured, Claude Pro's cap is about $155 of API value per week.
🇨🇳 Wenxin10 places · 3 notes
The same question, answered by Wenxin alone. The final list above combines all the AIs.
- 1Claude Pro20/𝑚𝑜(𝐴𝑛𝑡ℎ𝑟𝑜𝑝𝑖𝑐2026)
20 buys ~3.6 heavy agent-days a week — the clearest published meter at the low tier
Source ↗ - 2ChatGPT Plus 20/mo (OpenAI 2026)
20 with Codex agent: ~2.1 days before the fuzzy cap bites
Source ↗ - 3GitHub Copilot Individual 10/mo (GitHub 2026)
At 10 this is the cheapest visible meter — but it is code-only, not general agent
- 4Cursor Pro20/𝑚𝑜(𝐶𝑢𝑟𝑠𝑜𝑟2026)
20 gets you ~4 days of coding-agent fire — the best value at the low tier for coders
Source ↗ - 5The X Premium 30/mo (xAI 2026)
After the 1 Nov cut: 4× Plus pool, no published task count — you fly blind
- 6Claude Max 100/mo (Anthropic 2026)
100 opens Opus 5.5 all week — ~5.5 heavy days, the gold standard for clarity
Source ↗ - 7ChatGPT Pro 200/mo (OpenAI 2026)
200 for o3-level agent: ~4.8 days — but OpenAI keeps the meter behind a curtain
Source ↗ - 8The X Super100/𝑚𝑜(𝑥𝐴𝐼2026)
100 for a 15× pool with no task count — you pay more to guess less
- 9Gemini Advanced 20/mo (Google 2026)
20 with Gemini 2.5 Pro: ~2.8 days — Google's meter is clearer than X's but not as tight as Anthropic's
Source ↗ - 10The X Ultra 200/mo (xAI 2026)
200 for 50× Plus with zero published task count — the most expensive blind spot
What Wenxin noticed
Anthropic's 5-hour rolling window is the only published reset timer at the 20 tier — every other vendor hides or soft-caps their weekly allowance, making it impossible to plan agent work across a week.
The 1 November Grok Bot cut disproportionately hurt paying tiers: Premium users lost an estimated 40–60% of weekly messages while the multiplier labels (4×, 15×, 50×) stayed unchanged, creating a mismatch between advertised and actual value.
Cursor Pro at 20/mo delivers more usable coding-agent days than ChatGPT Plus at the same price because its 500 fast-request monthly pool is explicit and Cursor's agent loop consumes fewer tokens per task than a chat-based agent.
🇨🇳 Qwen10 places · 2 notes
The same question, answered by Qwen alone. The final list above combines all the AIs.
- 1Claude Max ($100)
A $100 five-hour rolling window that resets automatically, keeping heavy agent work predictable without a shared quota.
Source ↗ - 2GitHub Copilot Max ($100)
A $100 monthly credit pool that clearly prices premium model tokens instead of hiding behind an opaque request multiplier.
Source ↗ - 3The X Subscription Ultra ($200)
A $200 shared pool offering 50x the base allowance, but the lack of a published task count makes the Grok Bot cut unpredictable.
Source ↗ - 4ChatGPT Pro ($200)
A $200 tier that unlocks GPT-6, yet consolidates chat and Codex into a single shared quota that frustrates heavy coders.
Source ↗ - 5Cursor Pro ($20)
At $20, it offers extended agent limits, but third-party model usage is billed separately and drains fast on heavy CLI tasks.
Source ↗ - 6The X Subscription Super ($100)
A $100 tier with a 15x weekly pool multiplier that stretches further than Plus but shares the same unsized Grok Bot meter.
Source ↗ - 7Claude Pro ($20)
The $20 baseline for Claude Code users, relying on strict 5-hour rolling limits that prevent week-long runaway agent loops.
Source ↗ - 8Windsurf Pro ($20)
Priced at $20, it uses 'Flow Actions' to meter agentic usage, but users report burning through weekly credits on multi-file edits.
Source ↗ - 9ChatGPT Plus ($20)
The $20 standard plan includes Codex, but hitting the 5-hour window cap is a common complaint among casual agent runners.
Source ↗ - 10The X Subscription Premium ($30)
At $30, it grants a 4x multiplier over the free tier, the lowest tier to access the unified Grok Bot agent pool.
Source ↗
What Qwen noticed
The consolidation of ChatGPT and Codex into a single shared quota on the Pro tier has sparked significant backlash from developers who feel agent work cannibalizes general chat limits.
GitHub's shift to usage-based billing for Copilot Max replaces flat requests with a clear $100 monthly credit pool, making third-party premium model costs transparent.