PlanYield Get the Monday diff
Scoreboard / Glossary

AI coding plan glossary

Data as of · Vendor check: dated on each row · Page updated

Short answer (as of 18 Nov 2026): PlanYield’s glossary defines 36 terms that AI coding plans use for limits and billing: the 5-hour window and weekly cap, banked and global resets, usage multiples such as 5x Pro, GitHub Copilot AI credits at $0.01 each, Cursor’s usage pool, Kiro credits and multipliers, and the Fable cap. Each definition links the vendor source or the PlanYield page that uses it.

Cite this: PlanYield, 'AI coding plan glossary', updated 18 Nov 2026, https://planyield.dev/glossary/. CC BY 4.0.

5-hour window

A usage limit measured over five hours. On Claude Pro and Max it is a rolling session limit with a weekly limit on top, and Claude Ultra has none for verified accounts (claude.com/pricing). On Codex with ChatGPT Plus and Business Standard seats, a new window starts with your first message after the previous one ends, so it is not a rolling clock (OpenAI Help Center); Google AI Pro and Ultra refresh their Antigravity quota every five hours until a weekly limit is reached.

Weekly limit (weekly cap)

A cap on usage per week, on top of a shorter window or instead of one. Claude’s weekly limit resets at a fixed day and time assigned to your account, not from your first message (Anthropic help centre); ChatGPT Pro, Claude Ultra (verified accounts) and Devin Max have only a weekly limit. None of these vendors publishes the weekly limit as a fixed message count.

Banked reset

A usage reset saved to your account until you use it or it expires; using it refreshes your 5-hour and weekly Work and Codex limits and moves your weekly reset date (OpenAI Help Center). OpenAI’s 28-day Codex program had 3 banked resets: 8 Oct, 1 Nov, and 2 Nov.

Global reset

In OpenAI’s 28-day Codex program, a usage reset that was not banked: it was announced for that day, not saved to your account for later. 11 of the program’s 14 resets were global.

Saved limit reset (Claude)

A one-off reset Anthropic gave Claude subscribers on 23 Sep 2026 (SGT), used from Settings > Usage on the web or Claude Desktop. A @ClaudeDevs post said to use it before 22 Oct 2026.

Usage multiple (for example “5x Pro”)

A plan’s usage stated as a multiple of the vendor’s own base plan: Claude Max is 5x Claude Pro, Cursor Premium 4x Cursor Plus, and Google AI Ultra $100 5x Google AI Pro. Vendors publish the multiple, not the base amount, so a multiple does not convert across vendors.

AI credit (GitHub Copilot)

GitHub Copilot’s usage unit since 1 Jun 2026: one AI credit is $0.01. Paid plans include monthly credits (Pro 1,500, Pro+ 7,000, and Max 20,000) that reset at 00:00 UTC on the 1st and do not roll over; code completions are not billed in credits on paid plans (docs.github.com).

Base allotment (base credits)

The fixed part of a Copilot plan’s monthly AI credits. GitHub says base credits match the subscription price and do not change: Pro 1,000, Pro+ 3,900, and Max 10,000.

Flex allotment

The variable part of a Copilot plan’s monthly AI credits, on top of the base credits: Pro 500, Pro+ 3,100, and Max 10,000. GitHub describes it as variable, so it can change (docs.github.com).

Usage pool (Cursor)

The usage included in a Cursor plan, refreshed weekly. Plus is the 1x pool; Premium is 4x, Super 15x and Ultra 50x. On Premium and above, third-party models are billed at API list price outside the pool (cursor.com/pricing).

Top-ups (Cursor)

Prepaid credit bought past Cursor’s weekly pool, valid 365 days and spent at Cursor-model token rates, for example Grok 4.7 at $2 input and $6 output per 1M tokens (cursor.com).

Kiro credit

Kiro’s usage unit. Each plan includes monthly credits (Free 50, Pro 1,000, Pro+ 2,000, Pro Max 5,000, Power 10,000) that reset each billing month; unused plan credits do not roll over (kiro.dev).

Credit multiplier (Kiro)

How many Kiro credits a model uses compared with Auto at 1.0x. The Kiro Pro plan lists Claude Opus 5.5 at 2.0x, Claude Sonnet 5.5 at 1.3x and GPT-5.6 Sol, Terra and Luna at 4.4x, 2.2x and 0.6x (kiro.dev/docs/models).

Add-on credits (Kiro)

Prepaid Kiro credits on paid plans at $0.04 each, in packs from $5 (125 credits) to $100, up to 5 packs. Plan credits are used first, add-ons expire 12 months after purchase, and Free cannot buy them.

Credits (GLM Coding Plan)

Z.ai’s usage unit for the GLM Coding Plan. A call costs (input tokens × 6.9 + cached input × 1.7 + output × 24) ÷ 10,000 credits on GLM-5.3, and off-peak use costs half. Lite has 2,000 credits per 5 hours and 10,000 a week, Pro 12,000 and 60,000, and Max 28,000 and 140,000 (docs.z.ai).

Credits (Alibaba Token Plan)

Alibaba Cloud Model Studio’s unit for Token Plan usage; a call’s Credits depend on the model, tokens, thinking mode and tools. Token Plan includes 20,000, 45,000, 80,000, and 200,000 Credits a month on Lite, Essential, Standard, and Pro, reset each subscription month, with Extra Bundles past that (alibabacloud.com).

Rolling 5-hour window

A five-hour limit in which usage comes back five hours after it was used, a little at a time, instead of all at once at a fixed time. Kimi Code’s new plans have a rolling 5-hour rate window and no weekly limit; Alibaba’s Coding Plan releases each call 5 hours after use; Z.ai’s GLM Coding Plan refreshes its 5-hour credits 5 hours after use (vendor docs).

Extra Usage (Kimi Code)

Kimi’s balance for use past the plan quota: a prepaid RMB balance shared with Kimi on the web, priced close to Kimi Open Platform API rates, with an optional monthly spending cap. The balance does not expire, and only subscribers can turn it on (Kimi Code docs).

Request caps (Alibaba Coding Plan)

Alibaba’s Coding Plan counts model calls, not tokens. Pro allows 6,000 requests per 5 hours, 45,000 a week and 90,000 a month, all at once; when a cap is reached, calls fail until it resets, with no overage (alibabacloud.com).

Fable cap (Claude)

Anthropic’s limit on Fable models. On Max, Ultra and Team Premium seats Fable can use up to 50% of weekly limits; on Pro and Team Standard seats it sits outside the plan limits and runs on usage credits only.

Usage credits (Claude extra usage)

Pay-as-you-go usage past the Claude plan limits, at standard API rates. Prepaid bundles cost 10–30% less, up to $2,000 a month at the discount on Pro, Max and Ultra and $3,000 on Team (Anthropic help centre).

Monthly API credits

API credits some plans include each month: Claude Max $100 and Ultra $200 for the Claude Platform, and ChatGPT Pro $100, $200 or $500 by tier for the OpenAI API. They are for the vendor’s API, not for Claude Code or Codex usage, and they do not raise plan limits.

ChatGPT credits (Codex)

Credits ChatGPT Plus users can buy to keep using Codex past the limits, charged per 1M tokens by model (for example 50 credits per 1M input tokens on GPT-6.1 Sol). OpenAI does not state the dollar price of one credit.

Local messages (Codex)

OpenAI’s unit for the Codex 5-hour estimates on ChatGPT Plus, for example 15–160 GPT-6.1 Sol local messages per 5 hours. The ranges depend on task size.

Fast and Ultrafast modes (Codex)

Fast mode uses included usage at 2x for GPT-6.2 Astra/Sol and 2.5x for GPT-6.1 Sol/Terra and GPT-6 Astra/Sol/Luna; Ultrafast Pro 500: GPT-6.2 Astra and GPT-6.2 Sol (launch post); Ultrafast Pro 200: "bringing Ultrafast to Pro 200 subscribers, starting with GPT-6.2 Sol" (launch post); Per-model Ultrafast included usage: GPT-6.2 Astra and GPT-6.2 Sol at 6x Standard (pricing page); GPT-6 Astra and GPT-6.1 Sol Ultrafast on Pro 500 remain 8x included usage and 6x credits. The release note also says: "Pro $200 subscribers can use Ultrafast with Sol." The pricing page’s Pro card still lists "Ultrafast mode access on Pro $500". Eligible Enterprise and Edu workspaces support both GPT-6.2 models once an admin enables model access and Ultrafast separately.

Daily and weekly quota (Devin)

Devin Pro and Teams full seats have a daily allowance and a weekly quota; the daily allowance is more than one seventh of the weekly one, so weekend work does not cost weekly capacity. Devin Max is weekly only, and Cognition does not publish the sizes (docs.devin.ai).

On-demand credits (Devin)

Credits that keep Devin working past the quota, billed at API pricing. Cognition says purchased credits roll over and never expire.

Purchased AI credits (Google Antigravity)

Credits bought through Google One that Google AI Pro and Ultra can spend in Antigravity past their limits, at standard Gemini Enterprise consumption pricing when AI Credit Overages is set to Always. Bundle prices sit behind sign-in, and the credits are not sold in Japan.

Wrap-Up Allowance (Claude Code)

Anthropic’s allowance that may let Claude Code keep working briefly after you reach the 5-hour limit, to finish the step it is on. The usage counts toward the weekly limit; which versions and plans get it, and how often, are on Anthropic’s help page.

Legacy plan

A plan the vendor no longer sells to new customers but keeps for existing subscribers. Since 13 Oct 2026, Cursor’s old Pro ($20), Pro+ ($60), Ultra ($200) and Teams plans are legacy plans, and switching to a new plan is permanent.

As-of date (“Data as of”)

The date the PlanYield data was assembled (currently 18 Nov 2026). It is not a vendor check: each plan carries its own checked date, and a check more than 9 days older than the as-of date is flagged “re-check pending”.

Checked date

The day PlanYield read a plan on the vendor’s own pricing or docs page, printed next to its source. Prose that cites a check quotes this date, never the as-of date.

Verified (status label)

Read on the vendor’s own pricing or docs page, on the date printed with the plan as “checked”.

Partly verified (status label)

Read on a vendor page, but part of the figure is not published there or could not be read; the note next to the source says which part.

Secondary (status label)

Taken from a source that is not the vendor, such as news coverage or search results, and named on the plan or entry. Treat it with care.

Vendor post on X (status label)

A change-log entry sourced from a post on X by the vendor or a named member of its staff. It has a neutral colour because X posts can be edited or deleted.

Dataset licence: CC BY 4.0.