subscriptions on tomrochette.com

ChatGPT plans

tom@tomrochette.com (Tom Rochette) — Sat, 26 Sep 2026 00:00:00 +0000

ChatGPT plans are OpenAI’s consumer and team subscriptions, and every tier from Free upward now carries Codex, which makes them the subscription path to OpenAI’s coding models.

OpenAI publishes message ranges and credit rates where Anthropic publishes multipliers, but it spends the savings on complexity: six tiers, credits, speed multipliers, and a shared pool that Work, images, and voice all drain.

What it is
#

The 2026 lineup: Free ($0), Go ($8/month), Plus ($20/month), Pro from $100/month in 5x and 20x versions, Business ($25/user/month, $20 annual, minimum 2 seats), and Enterprise custom. Codex ships on web, CLI, IDE extension, and iOS, with cloud integrations (GitHub code review, Slack, Linear) from Plus up, and ChatGPT Work usage draws the same pool and rates as Codex. The GPT-6 lineup (Astra, Sol, Luna) carries the plans; GPT-5.5 retires from ChatGPT, Work, and Codex on all plans on October 14, 2026.

Status
#

Active, with the developer pricing page fetched and current as of 2026-09-26, and the widest subscription reach of any vendor here. The 2026 lineup changed materially: Business replaced the Team plan on April 2, the same day Codex billing moved from per-message to token-based credits, and the Go tier appeared below Plus. The current GPT-6 lineup (Astra, Sol, Luna) carries the plans, with Sol and Luna launched 2026-09-22 per the Model Selection guide and already listed as included on Plus and Pro in the fetched plan documentation.

Strengths
#

The most published mechanics in the category: official per-model message ranges per 5-hour window, for example GPT-6 Luna at 350-3,000 local messages on Plus versus 7,000-56,000 on Pro 20x.
Credits extend usage past included limits without upgrading, and Business sits at $25/user with SSO and no training on your data, the sensible team floor.
The Go tier at $8 is the cheapest named subscription carrying an agentic coder in this category.
Cloud tasks open pull requests end to end, the strongest hosted-agent story among the vendor plans.

Cautions
#

Everything shares one allowance: Codex, ChatGPT Work, image generation (3-5x faster burn), and desktop voice all draw the same pool, and cloud chats run GPT-5.6 Sol, which may consume more than local messages.
A practitioner’s note from real engagements: two engineers on one shared window exhausted it by early afternoon, the April switch to credits changed budgeting for everyone.
Real-world cost lands near $100-$200 per active developer per month once usage is heavy, which is Pro territory, not Plus.
Speed configurations and fast mode multiply credit burn, so the same prompt costs differently depending on settings you may not remember touching.
The consumer pricing page blocks automated fetches (403 this run), so plan verification runs through the developer docs.

Pricing
#

Free: $0, Codex for quick tasks with limited usage. Go: $8/month, lightweight coding. Plus: $20/month, a few focused sessions a week, GPT-6 Sol and Luna included. Pro: $100/month for 5x Plus usage, $200/month for 20x. Business: $25/user/month ($20 annual, minimum 2 seats), Codex via workspace credits. Enterprise: custom, flexible credit pricing. Credits purchase past included limits; API-key Codex usage bills at API rates outside any plan.

Price history
#

Date	Plan	Change	Source
2026-09-26	Free to Pro	Baseline: Free $0, Go $8, Plus $20, Pro $100 (5x) and $200 (20x) per month	https://developers.openai.com/codex/pricing
2026	Pro	Split into the $100 (5x) and $200 (20x) usage tiers; a single-tier Pro preceded it	https://automationatlas.io/answers/chatgpt-codex-pricing-explained-2026
2026-04-02	Business	Team at $30/user replaced by Business at $25/user ($20 annual, minimum 2 seats)	https://automationatlas.io/answers/chatgpt-codex-pricing-explained-2026
2026-04-02	Billing	Codex usage moved from per-message to token-based credit billing across Plus, Pro, Business, and Enterprise	https://automationatlas.io/answers/chatgpt-codex-pricing-explained-2026

Compared to
#

Claude plans: the mirror subscription at identical price points; OpenAI publishes ranges, Anthropic publishes multipliers, and the models themselves decide more than the mechanics.
GLM Coding Plan: the price-floor alternative, roughly one-seventh of Plus for the GLM line.
API billing (Model Provider Feature Matrix): per-token and uncapped, the right path for CI, and the escape hatch when shared windows keep interrupting real work.

Bottom line
#

Recommended for engineers inside the OpenAI ecosystem who want hosted agent features (cloud tasks, code review, Slack) with usage they can budget from published tables; Plus to start, Pro 5x when sessions interrupt. Not for predictable maximum throughput per dollar, where flat open-model plans win. My disagreeable claim: the published message ranges are more concrete than Anthropic’s multipliers but just as unusable in practice, because model choice, context size, and tool calls swing real consumption by an order of magnitude inside those ranges.

Changes
#

2026-09-26 - Created when the owner asked why the Claude and OpenAI subscriptions were missing from this category.

References
#

https://developers.openai.com/codex/pricing - plan prices, per-model message ranges, credit rates, GPT-5.5 retirement, feature availability matrix (fetched 200, 2026-09-26)
https://automationatlas.io/answers/chatgpt-codex-pricing-explained-2026 - the April 2 Business replacement, the credit-billing switch, Business seat pricing, the real-world cost range and shared-window practitioner note (fetched 200, updated 2026-07-31)
https://www.simplemetrics.xyz/chatgpt-codex-limits-2026 - independent limit analysis, the ranges-vs-fixed-caps framing, GPT-5.6 family tables (fetched via search extraction, updated 2026-09-09)
https://help.openai.com/en/articles/11481834-chatgpt-rate-card-business-enterpriseedu - the business credit rate card and the August 31 GPT-5.4 retirement note (linked from the fetched developer pricing page; not separately fetched)
https://chatgpt.com/pricing - the consumer plan page (403 to automated fetchers this run; plan prices verified through the developer docs instead)

Claude plans

tom@tomrochette.com (Tom Rochette) — Sat, 26 Sep 2026 00:00:00 +0000

Claude plans are Anthropic’s consumer subscriptions, Free, Pro, and Max, and from Pro upward they are the only way to run Claude Code without paying per token.

The subscription is a multiplier on an unpublished quota, not a token bucket: you are buying 1x, 5x, or 20x of a number Anthropic never publishes, drawn from one shared pool that chat, desktop, and Claude Code all drain together.

What it is
#

Three consumer tiers from Anthropic (PBC): Free at $0 with no Claude Code, Pro at $20 per month ($17 per month billed annually at $200 up front), and Max from $100 per month in 5x and 20x versions. Pro carries Claude Code, Claude Science, Design, Slides, Docs, and Projects; Max adds priority access at high-traffic times, higher output limits, and early access. Team seats run $25 per month ($20 annual) for Standard and $100 ($125 monthly, $100 annual) for Premium with 5x Standard usage; Enterprise is custom.

Status
#

Active and the default way engineers pay for Claude Code, with the pricing page fetched and current as of 2026-09-26. The 2026 record is lively: an April pricing-page test briefly removed Claude Code from Pro and was reverted within a day, a caching bug behind spring “usage drain” complaints was postmortemed on April 23 with limits reset, 5-hour limits were permanently doubled on May 6, and the weekly-limit promotion of May 13 was extended through August 19, 2026. The subscription has been repricing its value, not its price: the dollar figures are stable while the quota they buy keeps moving, which makes old guides the main hazard.

Strengths
#

Flat pricing against metered fear: one widely cited power user pushed roughly 10 billion tokens through eight months, about $15,000 at Opus API rates, on a $200 Max subscription (secondary writeup, hold loosely).
Usage credits continue a session past the plan limit at API rates under a cap you set, so a long task need not die at the wall.
Max gets priority access at peak times, which is when Pro users feel throttling.
The limits moved in your favor twice in spring 2026, permanently for the 5-hour clock.

Cautions
#

One shared pool: a morning of claude.ai chat shrinks the afternoon Claude Code session, and agent teams burn about 7x a normal session in plan mode, per Anthropic’s own costs docs.
Fable 5.1, the flagship, is capped at 50% of weekly limits on Max and is available on Pro only through usage credits, the first tightening of 2026 (July 20).
Anthropic does not publish token quotas for any plan; multipliers are all you get, and third-party figures like “220,000 tokens per 5 hours” are estimates.
The weekly-limit promotion lapsed on paper on August 19, 2026, so weekly walls may now arrive about a third sooner than spring habits suggest.
A June 15 plan to move programmatic usage (Agent SDK, headless runs) to separate credit pools was paused the day it was due and never shipped; watch for it to return.

Pricing
#

Free: $0, no Claude Code. Pro: $20/month, $17/month billed annually ($200 up front), Claude Code included. Max 5x: $100/month, 5x Pro usage. Max 20x: $200/month, 20x Pro usage, monthly billing only. Team Standard $25/seat/month ($20 annual), Team Premium $100/seat/month ($100 annual), Enterprise custom; usage credits extend any plan at API rates under a spending cap.

Price history
#

Date	Plan	Change	Source
2026-09-26	Pro / Max	Baseline: Pro $20/month ($17 annual), Max 5x $100/month, Max 20x $200/month, Fable on Pro via usage credits and on Max at 50% of weekly limits	https://claude.com/pricing
2026-05-06	Usage	5-hour limits permanently doubled for Pro, Max, Team, and seat-based Enterprise; peak-hour throttling removed	https://ccforeveryone.com/guides/claude-code-limits-and-pricing
2026-05-13	Usage	Weekly limits raised 50% as a promotion, extended three times through 2026-08-19	https://ccforeveryone.com/guides/claude-code-limits-and-pricing
2026-07-20	Fable	Fable 5.1 added to Max, Team Premium, and Enterprise up to 50% of weekly limits; Pro and Team Standard only via usage credits, the year’s first tightening	https://ccforeveryone.com/guides/claude-code-limits-and-pricing
2026-07-25	Models	Opus 5 released on all paid plans at unchanged subscription prices, default on Max	https://ccforeveryone.com/guides/claude-code-limits-and-pricing

Compared to
#

GLM Coding Plan: the price-floor challenger at $18, with published credit mechanics and China-based routing; Claude plans cost more and publish less, and buy the stronger frontier model.
ChatGPT plans: the direct rival subscription, which publishes per-model message ranges and credit rates where Anthropic publishes multipliers only.
API billing (Model Provider Feature Matrix): still the right path for CI and automation, and the comparison that makes Max look cheap only works at genuinely heavy usage.

Bottom line
#

Recommended for engineers whose daily driver is Claude Code and who can live with a shared pool and unpublished quotas; start at Pro and upgrade on the second real weekly cap. Not for automation-heavy workloads, where the API or a flat open-model plan fits better. My disagreeable claim: most Max subscribers are paying for a productivity feeling rather than a quota, because the median engineer never hits the Pro weekly wall that justifies 5x.

Changes
#

2026-09-26 - Created when the owner asked why the Claude and OpenAI subscriptions were missing from this category.

References
#

https://claude.com/pricing - plan prices, annual discount, Claude Code inclusion matrix, Fable access rules (fetched 200, 2026-09-26)
https://ccforeveryone.com/guides/claude-code-limits-and-pricing - the 2026 change log (May 6 doubling, May 13 promotion, July 20 Fable tightening, paused Agent SDK split), limit mechanics, Team seat prices, the API-versus-subscription math (fetched 200, updated 2026-08-05)
https://www.ssdnodes.com/blog/claude-code-pricing-in-2026-every-plan-explained-pro-max-api-teams/ - independent plan walkthrough, the instrumented-usage cost comparison it cites (fetched via search extraction, 2026-09-26)
https://www.layer3labs.io/guides/claude-pro-vs-max-for-teams - the weekly-cap skepticism and the upgrade heuristics (fetched via search extraction, 2026-09-26)
https://support.claude.com/en/articles/11049741-what-is-the-max-plan - the Max plan article: $100/$200 tiers, monthly-only billing, 5-hour and weekly limit mechanics, the discretionary-limits reservation (fetched 200, 2026-09-26)

Google AI plans

tom@tomrochette.com (Tom Rochette) — Sat, 26 Sep 2026 00:00:00 +0000

Google AI plans are Google’s consumer subscription ladder, Google AI Plus, Google AI Pro, and Google AI Ultra in 5x and 20x versions, and from Plus upward they are how individuals raise the limits on Antigravity, AI Studio, Jules, and the Gemini app.

The plans sell multipliers on unpublished quotas plus a credit pool billed at API consumption rates, and Google’s own free Antigravity tier is good enough that most engineers hitting a wall should check whether they need to pay at all.

What it is
#

Three consumer tiers sold through Google One: Plus (2x Gemini access versus non-subscribers, 400 GB storage), Pro (4x, 5 TB, $10 in monthly Google Cloud credits), and Ultra in 5x (5x Pro, 20 TB, $40 credits) and 20x (20x Pro, 30 TB, $100 credits) versions, as of 2026-09-26. The plans carry the Gemini app, Google Flow, Gemini Notebook, AI Studio, Android Studio, Chrome auto browse, and expanded Antigravity and Jules limits, with Flow credits of 200 to 25,000 per month across the ladder. Personal Google Accounts only: Workspace customers are pushed to Gemini add-ons, and organizations to Google Cloud terms with consumption-based API pricing. The FAQ confirms the ladder’s churn: the old Google AI Premium plan was renamed Google AI Plus.

Status
#

Active and restructured repeatedly, with the tiers, multipliers, and storage allotments verified from Google’s own pages on 2026-09-26. Plus is available in over 160 countries, Pro and Ultra in over 150. The 2026 record shows a rename (AI Premium to Plus) and an Ultra split into 5x and 20x columns, both visible on the current comparison table. Prices are the weak point of the product surface: the marketing pages render the dollar figure client-side per region, so my US and Canada fetches returned the ladder with the price stripped, and third-party guides are the only place the US numbers appear in text.

Strengths
#

The free Antigravity tier it upgrades is real: unlimited tab completions and command requests, a weekly baseline quota, and access to Gemini, Claude, and open-weight models.
The flexible AI credit pool converts a hard wall into metered overage at documented consumption pricing, with a Never/Always overage setting instead of a surprise bill.
Ultra refreshes its quota every five hours against the free tier’s weekly clock, the difference that matters for daily-driver agent work.
The bundled Google Cloud credits ($10, $40, $100 monthly by tier) partially self-fund Ultra for anyone who also calls the Gemini API.

Cautions
#

Google publishes no quota numbers for any tier, and the Antigravity docs say limits are “correlated with the amount of work done by the agent”, so one gnarly prompt can cost a day of quota.
The AI credits purchase page sits behind Google sign-in, so credit prices are the least documented numbers in the whole stack.
Google’s own pages disagree on Plus storage, 400 GB on the AI plans page versus 2 TB on the Canadian plans page, both fetched 2026-09-26.
The consumption rates your credits burn at double on January 1, 2027, when the current Gemini API promotional prices expire.
Tier mechanics have already been changed out from under users once: the May 2026 “Antigravity bait and switch” thread drew 771 points.
No bring-your-own-key and no organizational contract tiers on the consumer path.

Pricing
#

Canadian storefront, fetched 2026-09-26: Google AI Plus CA$13.99/month, Google AI Pro CA$26.99/month, Ultra not listed on that page. US dollar list prices did not render in my fetched HTML; a third-party comparison table lists Plus around $8/month as the cheapest paid path and Ultra 20x at $200/month as the top individual tier (blocked fetch, see references). Organizations: Google Cloud terms with consumption-based API pricing, included in select Gemini Enterprise subscriptions. Credits bill at standard Gemini Enterprise consumption rates, anchored by the Gemini API price list: Gemini 3.8 Flash at $0.75/$3.75 per million tokens through December 31, 2026, doubling in 2027, and Gemini 3.1 Pro at $2.00/$12.00.

Price history
#

Date	Plan	Change	Source
2026-08-24	Individuals / Pro / Ultra	Baseline: Antigravity free at $0 with basic weekly rate limits, Google AI Pro and Ultra the paid paths that raise limits and add the AI credit pool	https://antigravity.google/pricing
2026-09-26	Ladder	Verified: Plus / Pro / Ultra (5x, 20x) ladder at 2x/4x/5x-Pro/20x-Pro access, AI Premium confirmed renamed to Google AI Plus, CA$13.99 Plus and CA$26.99 Pro on the Canadian storefront	https://one.google.com/about/google-ai-plans

Compared to
#

Claude plans: the same multiplier-on-a-secret-quota pattern, but Google bundles harder (Flow credits, cloud credits, YouTube) and prices its entry tier lower.
ChatGPT plans: the rival ladder, which publishes per-model message ranges where Google publishes adjectives.
Google Antigravity: the surface these plans meter; its free tier undercuts the paid ladder for light use, which is the sharpest critique of the ladder itself.
Model Provider Feature Matrix: the API-side alternative, still the right answer for automation and CI, and the baseline your credit pool burns at.

Bottom line
#

Recommended for engineers who have actually hit the free Antigravity weekly wall and want metered overage at known API rates; start at Pro, and only after two real capped weeks. Not for light users, who should stay on the free tier, and not for automation-heavy workloads, which belong on the API. My disagreeable claim: Ultra is priced for video generation more than for coding, because the 25,000 monthly Flow credits are its most concrete differentiator, so engineers buying Ultra 20x for Antigravity are mostly buying Flow.

Changes
#

2026-09-26 - Created when the owner asked for any remaining subscription providers.

References
#

https://one.google.com/about/google-ai-plans - the plan ladder, comparison table (multipliers, Flow credit quotas, Antigravity and Jules rows), AI Premium rename, country counts (fetched 200, 2026-09-26)
https://one.google.com/intl/en_us/about/google-ai-plans/ - US variant: $10/$40/$100 monthly Google Cloud credits by tier, Home Premium bundle values, Ultra 5x and 20x storage tiers (fetched 200, 2026-09-26)
https://one.google.com/about/plans - Canadian storefront prices, CA$13.99 Plus and CA$26.99 Pro, no Ultra listed (fetched 200, 2026-09-26)
https://antigravity.google/pricing - the $0 individual tier, Pro/Ultra as the paid paths adding rate limits and the AI credit pool, Google Cloud consumption terms for organizations (fetched 200, 2026-09-26)
https://antigravity.google/docs/plans - quota mechanics: five-hour Ultra refresh, weekly free quota, credits billed at standard Gemini Enterprise consumption pricing, Never/Always overages, no BYOK (fetched 200, 2026-09-26)
https://ai.google.dev/gemini-api/docs/pricing - the consumption-rate anchor: 3.8 Flash $0.75/$3.75 per million tokens until 2026-12-31 then doubled, 3.1 Pro $2.00/$12.00 (fetched 200, 2026-09-26)
https://news.ycombinator.com/item?id=48222529 - critical source: the May 2026 “Antigravity bait and switch” thread, 771 points, on tiers changing under existing users (fetched 200, 2026-09-26)
https://codeagentswarm.com/en/guides/grok-build-pricing - third-party table naming Plus around $8/month and Ultra 20x at $200/month; fetched 429 this run, figures carried from the 2026-08-25 verification, hold loosely

Qwen Coding Plan

tom@tomrochette.com (Tom Rochette) — Sat, 26 Sep 2026 00:00:00 +0000

The Qwen Coding Plan is Alibaba Cloud Model Studio’s flat-rate subscription that sells Qwen plus rival models (Kimi, GLM, MiniMax) into coding agents through one plan-specific key. It is the only mainstream coding subscription I found that bundles competitors’ models under one fee, but it carries the strictest automation ban in this category and its product line was restructured twice in six months.

What it is
#

A subscription from Alibaba Cloud (Model Studio, the international DashScope platform) now sold in two generations. The older request-based Coding Plan charges $50/month for Pro and deducts per request; the newer credits-based Token Plan (Singapore region) deducts from a shared Credits pool spanning text, image, video, and speech models. Both expose OpenAI-compatible and Anthropic-compatible endpoints and require a plan-specific sk-sp- key with a dedicated base URL, so one subscription feeds Claude Code, OpenCode, Codex, Cursor, Cline, Qwen Code, Qoder, Kilo CLI, and OpenClaw. The Pro roster includes third-party flagships (kimi-k2.5, glm-5, MiniMax-M2.5) alongside qwen3.7-plus, qwen3.6-plus, and the qwen3-coder line.

Status
#

Actively sold but mid-transition, and the transition is not cosmetic. Coding Plan Lite closed to new subscriptions on 2026-03-20 and to renewals and upgrades on 2026-04-13. The official Token Plan FAQ states that Coding Plan Pro was a limited-quantity offering that is “no longer available once sold out”, recommends Token Plan instead, and provides no migration or upgrade path between the two products. As of 2026-09-26 the coding-plan docs (updated 2026-09-11) still headline the $50 Pro plan while the Token Plan guide (updated 2026-09-25) sits above it in the docs nav, and Token Plan is Singapore-region only.

Strengths
#

The multi-vendor roster is the real product. One key routes to Qwen, Kimi-K2.5, GLM-5, MiniMax-M2.5, and on Token Plan DeepSeek-V4-Pro plus the image, audio, and video families.
Token Plan entry is the cheapest on-ramp I verified in this category: $6/month limited-time ($8 list) for 11,500 Credits.
$50 Coding Pro bought up to 90,000 requests per month, 6,000 per 5 hours, 45,000 per week, which undercuts pay-as-you-go for anyone spending over $100/month on these APIs.
Fixed monthly billing with hard caps: when a window empties, calls pause instead of drawing on your account balance.

Cautions
#

The terms ban automated scripts, CI/CD, batch processing, and application backends, and violations “may result in subscription suspension or API key revocation”.
Critical coverage documents enforcement as automated and unpredictable: users report suspensions for rapid-fire long sessions and background-request tools, immediate loss of access for the rest of the billing period, and slow, opaque appeals; as the review puts it, “the suspension risk is not theoretical”.
Quota counts requests, not tokens: Alibaba’s own docs say simple tasks burn 5-10 model calls and complex ones 10-30 or more, so 90,000 requests is far fewer agent runs than it sounds.
Mixing the plan key with the general pay-as-you-go key produces unexpected API charges, a footgun the official FAQ dedicates a section to.
Pro models are pinned to exact snapshots and third-party reviewers report the roster changing with little notice, so a subscription can silently change what it buys.
Platform throttling under load is contractually “not considered a service interruption or a breach of contract”, so slowdowns carry no recourse.

Pricing
#

As of 2026-09-26 two price sheets coexist. Coding Plan Pro: $50/month, capped at 6,000 requests per 5 hours, 45,000 per week, and 90,000 per month, whichever hits first (official docs, updated 2026-09-11). Token Plan Personal (Singapore region): Lite $6 (list $8), Essential $10 (list $16), Standard $18 (list $25), Pro $68 (list $80) per month for 11,500/25,500/45,000/180,000 Credits, with team seats at $20/$75/$200 and extra bundles at $15 per 20,000 Credits. A third-party tracker (data updated 2026-09-24) reports China-side early-bird pricing of ¥39/¥139/¥499 per month, a limited-time night rate of 40% of normal Credits from 22:00, and 88% over-limit billing.

Price history
#

Date	Plan	Change	Source
2026-03	Lite / Pro	Baseline: Lite listed at ¥40/month in China and promoted as low as $3/month internationally, Pro at $50/month	codingplan.org; vibecoding.app
2026-03-20	Lite	Discontinued for new subscriptions	Alibaba Cloud coding-plan docs
2026-04-13	Lite	Renewals and upgrades discontinued	Alibaba Cloud coding-plan docs
2026-08	Token Plan	Request-based Coding Plan upgraded to a credits-based Token Plan with China early-bird ¥39/¥139/¥499 per month	codingplan.org
2026-09-26	Token Plan (intl)	Lite $6/$8, Essential $10/$16, Standard $18/$25, Pro $68/$80 per month; team seats $20/$75/$200	Alibaba Cloud Token Plan overview

Compared to
#

- GLM Coding Plan wins on quota per dollar; choose Qwen when one subscription covering several vendors’ models matters more.
- MiniMax Coding Plan follows the same China flat-fee pattern with a narrower roster and similar interactivity terms.
- OpenRouter sells the opposite contract: pay per token, no interactivity restrictions, no suspension risk.

Bottom line
#

Recommended for heavy interactive users inside supported harnesses who want Qwen and its rivals under one flat fee and can tolerate mid-cycle throttling. Not for CI/CD, batch jobs, unattended agent farms, or anyone who cannot absorb losing access mid-billing-cycle with no guaranteed refund. Claim to disagree with: the $50 Coding Plan Pro that Alibaba’s docs still headline is now the wrong purchase for almost everyone, because Alibaba’s own FAQ calls it a limited offering that is gone once sold out and steers new buyers to the credits-based Token Plan.

Changes
#

2026-09-26 - Created when the owner asked for any remaining subscription providers.

References
#

https://www.alibabacloud.com/help/en/model-studio/coding-plan - official Coding Plan docs: $50 Pro, 6,000/45,000/90,000 request caps, Lite dates of 2026-03-20 and 2026-04-13, automation ban, snapshot-pinned roster, sk-sp- key rules (fetched, HTTP 200).
https://www.alibabacloud.com/help/en/model-studio/token-plan-overview - official Token Plan overview: USD tiers and Credits, Singapore-only availability, and the FAQ stating Coding Plan Pro was limited-quantity with no migration path (fetched, HTTP 200).
https://www.alibabacloud.com/en/campaign/ai-scene-coding - official campaign and purchase page: limited-time USD prices, 2X credit promotion, Qwen3.8-Max positioning (fetched, HTTP 200).
https://www.alibabacloud.com/help/en/model-studio/token-plan-guide - official docs hub placing Token Plan above Coding Plan in the product nav as of 2026-09-25 (fetched, HTTP 200).
https://codingplan.org/en/plans/qwen - tracker: credits restructure, ¥39/¥139/¥499 early-bird, 40% night rate from 22:00, legacy ¥40/¥200 China prices (fetched, HTTP 200, data updated 2026-09-24).
https://vibecoding.app/blog/alibaba-coding-plan-review - critical review: suspension reports, opaque appeals, no rollover, roster changes without notice (fetched, HTTP 200, updated 2026-06-17).
https://common-buy-intl.alibabacloud.com/coding-plan - official purchase URL; fetch reached a 2FA verification wall, so no plan data was read from it (fetch blocked, disclosed).

SuperGrok

tom@tomrochette.com (Tom Rochette) — Sat, 26 Sep 2026 00:00:00 +0000

SuperGrok is xAI’s (SpaceXAI LLC’s) consumer subscription ladder for Grok, running from a free tier to $300 per month, which now doubles as the way engineers pay for the Grok Build coding agent.

One subscription, one weekly pool: your Grok chat, your Grok Build runs, and your API calls all drain the same weekly bucket, so the plan you buy for coding is also the plan you spend by chatting.

What it is
#

A five-step consumer ladder sold by xAI for its Grok products on web, iOS, and Android: Free, SuperGrok Lite, SuperGrok, SuperGrok Plus, and SuperGrok Heavy, with Business and Enterprise seats above. The tiers meter chat, Imagine image and video generation, Voice, Grok Bot, and Grok Build, xAI’s open-source terminal coding agent, which carries its own row on the official plan comparison for every tier from Free to Enterprise, and the API pricing page separately refers to “Grok Build’s free tier”. Above the ladder sits the pay-per-token API, billed separately from any subscription. The official pricing page lists dollar figures only for Free, SuperGrok, and Plus; Lite and Heavy show as plan columns without public prices on the web card, so their numbers come from press coverage and third-party guides.

Status
#

Active and still being restructured: Lite was announced on 2026-03-25 at $10 per month while in testing with selected users, and SuperGrok Plus is absent from a complete tier guide dated 2026-07-06 yet present on the official page on 2026-09-26. The defining mechanic is recent and explicit: integration documentation states that every paid Grok subscription gets one weekly usage pool shared across Grok chat, Grok Build, and API access, and that paid usage pauses when the pool empties until the reset time shown in the Usage tab. Third-party harnesses can spend the same pool: Warp connects over OAuth and routes Grok models through your xAI account, with its requests labeled API in xAI’s usage dashboard. The ladder is converging on one fuzzy weekly pool instead of countable messages, while the tier list itself is still moving underneath it.

Strengths
#

The weekly pool is flexible rather than siloed, so a coding-heavy week can spend the budget on Grok Build and a research-heavy week on DeepSearch and chat.
Grok Build is included from Free upward, which makes the $0 tier a working entry point to a frontier coding agent, with paid tiers raising its limits from the same pool.
The API escape hatch is cheap for a frontier lab: grok-4.6 and grok-4.7 bill $2.00 input and $6.00 output per million tokens under 200k prompt, with cached input at $0.50, as of 2026-09-26.
The subscription travels outside xAI’s own apps, since Warp and similar tools authenticate against it directly.
Lite at $10 per month would be the cheapest paid on-ramp of any frontier-lab subscription in this category, once it ships generally.

Cautions
#

No quota is published anywhere: no messages, no tokens, no pool size, so the only way to size a plan is to burn a month and read the Usage tab.
The shared pool cuts both ways: an afternoon of video generation or idle chat quietly deletes tomorrow’s coding budget.
Lite is not confirmed as generally available, announced as a test with selected users in March 2026, and Heavy’s $300 figure rests on third-party guides because the official web card does not expose it.
Coding is the ladder’s weakest argument: the critical guide I fetched concludes Claude and ChatGPT lead on coding help and recommends SuperGrok for live X data and image work, not for code.
Imagine, the media product the higher tiers are priced around, is the surface under active lawsuits, a US congressional inquiry, and EU, UK, and Canada probes, per the same critical guide.
A subscription connected through a third-party harness carries no zero-data-retention guarantee from that harness, since retention on xAI’s side is governed by your own xAI account.

Pricing
#

Free: $0, limited usage, Grok Build included with limits. SuperGrok Lite: $10/month, announced 2026-03-25, basic creation tools, 480p video up to 6 seconds, 2x longer chats than Free, one AI agent, in testing at announcement. SuperGrok: $30/month, Grok 4.6, higher rate limits across all features, image and video generation. SuperGrok Plus: $100/month, everything in SuperGrok plus 1080p video, significantly higher usage across Chat, Imagine, Voice, and Build, priority access at peak times. SuperGrok Heavy: $300/month per third-party guides as of 2026-07-06, the multi-agent Grok 4 Heavy tier. API escape hatch, billed separately: grok-4.6 and grok-4.7 at $2.00/$6.00 per 1M tokens under 200k prompt ($4.00/$12.00 above), $0.50 cached input; grok-build-0.1 at $1.00/$2.00; Grok 4.7 Fast at 2x rates, exclusive to Cursor and Grok Build and excluded from Grok Build’s free tier.

Price history
#

Date	Plan	Change	Source
2026-01-24	SuperGrok / Heavy	Baseline: SuperGrok $30/month and Heavy $300/month, no Lite or Plus tier; the July 2026 update describes these two anchor prices as unchanged	https://aitoolanalysis.com/supergrok-subscription-price-2026/
2026-03-25	SuperGrok Lite	Announced at $10/month by Musk on X, in testing with selected users, global rollout promised later in the year	https://www.businesstoday.in/technology/story/xai-makes-grok-affordable-with-new-supergrok-lite-plan-check-price-and-what-it-offers-522462-2026-03-26
2026-09-26	SuperGrok Plus	Observed at $100/month on the official pricing page, absent from the 2026-07-06 complete tier guide, so added in the July to September window	https://x.ai/pricing

Compared to
#

Claude plans: the $20 to $200 Anthropic ladder that runs Claude Code; it matches SuperGrok at the $100 rung and carries the stronger coding reputation, while publishing no more about its quotas than xAI does.
ChatGPT plans: the rival consumer ladder at $20 to $200, which publishes per-model message ranges, a discipline xAI has not adopted.
The Model Provider Feature Matrix: the per-token path, and the reason the shared pool matters less if your coding runs on the API anyway.

Bottom line
#

Recommended for engineers already inside the Grok and X ecosystem who want one bill across chat and Grok Build and can tolerate an unpublished weekly pool. Not for coding-first engineers choosing a primary agent subscription, where Claude plans or the GLM Coding Plan buy more verified coding throughput per dollar. My disagreeable claim: SuperGrok Plus at $100 is a media-creator tier wearing an engineer’s price tag, because on a shared pool the extra $70 mostly buys 1080p video and peak-hour priority, not more code.

Changes
#

2026-09-26 - Created when the owner asked for any remaining subscription providers.

References
#

https://x.ai/pricing - official plan cards (Free $0, SuperGrok $30, Plus $100) and the plan comparison carrying Lite, Heavy, and Grok Build rows (fetched 200, 2026-09-26)
https://docs.warp.dev/agents/inference/grok-subscription - the one weekly pool shared across Grok chat, Grok Build, and API access, pause-on-empty behavior, third-party-harness ZDR caveat (fetched 200, page updated 2026-09-24)
https://docs.x.ai/developers/pricing - API escape hatch: grok-4.6/4.7 at $2/$6 under 200k prompt, $0.50 cached, $4/$12 above, grok-build-0.1 at $1/$2, Grok 4.7 Fast exclusivity (fetched 200, 2026-09-26)
https://www.businesstoday.in/technology/story/xai-makes-grok-affordable-with-new-supergrok-lite-plan-check-price-and-what-it-offers-522462-2026-03-26 - Lite announcement: $10/month, testing phase with selected users, 480p 6-second video, 2x chats, one agent (fetched 200, 2026-09-26)
https://aitoolanalysis.com/supergrok-subscription-price-2026/ - critical guide: full tier table including Heavy $300, the coding-benchmarks caveat, and the safety litigation section (fetched 200, updated 2026-07-06)
https://codeagentswarm.com/en/guides/grok-build-pricing - independent Grok Build plan-table guide, attempted four times this run and rate-limited (429) on every attempt, so nothing here is cited from it

subscriptions on tomrochette.com

ChatGPT plans

What it is #

Status #

Strengths #

Cautions #

Pricing #

Price history #

Compared to #

Bottom line #

Changes #

See also #

References #

Claude plans

What it is #

Status #

Strengths #

Cautions #

Pricing #

Price history #

Compared to #

Bottom line #

Changes #

See also #

References #

Google AI plans

What it is #

Status #

Strengths #

Cautions #

Pricing #

Price history #

Compared to #

Bottom line #

Changes #

See also #

References #

Qwen Coding Plan

What it is #

Status #

Strengths #

Cautions #

Pricing #

Price history #

Compared to #

Bottom line #

Changes #

See also #

References #

SuperGrok

What it is #

Status #

Strengths #

Cautions #

Pricing #

Price history #

Compared to #

Bottom line #

Changes #

See also #

References #

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Price history
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#