Changes that can move the estimates
Codex and Claude plans, their models and this site's method all change. These are the changes that can shift the weekly limit estimates, and changes to what a plan includes, newest first, each with the source it was checked against.
The shared-report API pools several recent weeks, so policy changes reach its estimates gradually as older weeks age out. Homepage category estimates change when a new selected run is published; higher-tier figures use stated assumptions. Daily Speed points record the fixed task on their UTC date. A change known only from second-hand reports stays off this list until its source can be read.
- · OpenAI
Pro allowances double, including the lower Pro $200 allowance
OpenAI says Pro 100 and Pro 500 now include twice their previous Work and Codex allowance. On Pro 200 only the lower allowance, the one new subscriptions got from September 29, doubled. Existing subscribers get the update without waiting for renewal, and Plus usage is unchanged. This entry previously recorded grandfathered Pro $200 subscriptions moving to the lower allowance after October 29. The homepage now multiplies the Plus estimate and API value by 10, 20 and 50 for Pro $100, $200 and $500, up from 5, 10 and 25.
Source: OpenAI: About ChatGPT Pro tiers
- · OpenAI
ChatGPT Pro includes monthly API credits
ChatGPT Pro includes $100, $200 or $500 a month in OpenAI API credits, matching the plan price, after linking an API organization you own. The subscription must be paid and at least seven days old. OpenAI excludes Codex usage through API keys and says the credits do not increase ChatGPT or signed-in Codex limits. This leaves the weekly limit and API-value estimates unchanged.
Sources: API credits for ChatGPT subscribers, OpenAI API changelog, Oct 28, 2026
- · OpenAI
GPT-5.5 retires from Codex
Announced on September 14: GPT-5.5 leaves ChatGPT, ChatGPT Work and Codex on all plans. Weeks spent on it give way to weeks on other models, and each model draws on a limit at its own rate.
Sources: OpenAI changelog, Sep 14, 2026, OpenAI pricing
- · tokenusage.dev
Homepage category cards and daily Speed use local runs
The homepage keeps its plan cards in seven category sections and fills them from selected local ChatGPT Plus and Claude Pro measurements, published through owner ingest and Firestore. Weekly capacity is projected from the overall weekly meter; higher-tier capacity and API value use scaling assumptions explained on the method page. Cards show weekly capacity and API value; the Response column is removed. Speed appears only in the daily chart from a separate fixed-task probe, with the former sharer-derived history archived privately. Missing results stay empty. The shared-report subscription API’s cell calculations, leaderboards and profiles retain their existing calculations.
Source: Homepage benchmark method
- · tokenusage.dev
Claude Code shares reach the Claude cells
The Claude Code edition of token-counter shares Claude Code usage as client claude-usage. Each weekly window counts the tokens between its first recorded percentage reading and the earliest reading at its peak, with a rejected weekly 429 taken as a 100% reading. Once enough sharers and windows stand behind a cell, their weeks move the Claude medians beside the restored reference dataset.
- · Anthropic
Max plans include monthly API credits
Max and Team plans now include monthly API credits for the Claude Platform: $100 a month on Max 5x and $200 on Max 20x. Pro is not eligible. Anthropic says the credits don't change usage limits in Claude or Claude Code and don't cover interactive Claude Code, so this change does not alter the weekly limit estimates. The homepage's API-value estimates price projected token usage and do not include these credits.
Sources: Claude release notes, Oct 7, 2026, Monthly API credits for Max and Team plans
- · OpenAI
GPT-6.1 Sol reaches Codex and becomes the CLI's default
OpenAI added GPT-6.1 Sol to Codex and ChatGPT Work, and Codex CLI 0.159.1 made it the default in its bundled model catalog. Sharers who keep the default move onto it, so weeks after this may hold a different number of tokens than weeks before.
Source: OpenAI changelog, Sep 29, 2026
- · OpenAI
GPT-6 Sol and GPT-6 Luna reach Codex
Rolling out to Plus and Pro, among others, at lower token prices than the GPT-5.6 models before them. OpenAI's pricing page says credit prices alone do not set how fast a plan's included usage goes, so the effect on a weekly limit is not stated. Weeks on the new models may hold a different number of tokens than weeks before.
Sources: OpenAI changelog, Sep 22, 2026, OpenAI pricing
- · OpenAI
GPT-5.3-Codex-Spark is deprecated
When its research preview opened to Pro on February 12, OpenAI said Spark's usage had limits of its own and did not count against the standard Codex limits. The plugin counts its tokens like any other, so a week with Spark in it can read high: more tokens than the weekly limit held. It is no longer available in the app, CLI or IDE extension.
Sources: OpenAI changelog, Sep 14, 2026, OpenAI changelog, Feb 12, 2026
- · Anthropic
Claude Code weekly limits settle 25% above their pre-promotion level
From May 13 through September 13, a promotion raised Claude Code's weekly limits by 50% on Pro, Max and Team plans, plus legacy seat-based Enterprise plans. From September 14, Anthropic kept part of it: weekly limits are 25% higher than before the promotion. A week during the promotion had a 20% larger allowance than a week after it (150 against 125).
- · OpenAI
GPT-5.4 and GPT-5.4 mini leave Codex for ChatGPT sign-in
OpenAI named GPT-5.6 Terra and GPT-5.6 Luna as their replacements. The old models stay available with an API key, which is billed at API rates rather than drawing on a plan's limits.
Sources: OpenAI changelog, Jul 31, 2026, OpenAI pricing
How the numbers work explains how the weekly estimates are made.