AI limit resets and limit changes since Sep 1, 2026, by provider
Counts are rows in our resets log plus our rate-limit and quota changelog entries, Sep 1, 2026 to Nov 3, 2026 (UTC dates). Method
OpenAI accounts for 11 of 12 resets, all 11 from one staff member’s posts on X, plus 3 limit changes, including the Oct 29 Pro 200 reversal¹.
| Share | ||||
|---|---|---|---|---|
| OpenAI (ChatGPT) | 11 | 3 | 14 | |
| Fireworks AI | 0 | 4 | 4 | |
| Amazon Bedrock (AWS) | 0 | 2 | 2 | |
| Anthropic (Claude) | 1 | 1 | 2 | |
| xAI (Grok) | 0 | 2 | 2 | |
| GitHub Copilot | 0 | 1 | 1 | |
| Hugging Face | 0 | 1 | 1 | |
| Together AI | 0 | 1 | 1 | |
| All providers | 12 | 15 | 27 |
The rows behind each count
OpenAI (ChatGPT) 14
- 2026-09-26Usage limits, all paid users across Codex and ChatGPT Work, after a disruptionsource
- 2026-10-06API usage tiers simplified from five to three (Build, Launch, Grow).source
- 2026-10-07Banked reset loaded into paid accountssource
- 2026-10-10Global reset by EOD (eligible users/limits not stated)source
- 2026-10-13Global reset by EOD PST (eligible users/limits not stated)²source
- 2026-10-15A reset (scope not stated in the post)source
- 2026-10-17A reset (scope not stated in the post)source
- 2026-10-20A reset (scope not stated in the post)source
- 2026-10-22A reset (scope not stated in the post)source
- 2026-10-23A reset (scope not stated in the post)source
- 2026-10-25A reset (scope not stated in the post)source
- 2026-10-26A reset (scope not stated in the post)source
- 2026-10-29OpenAI’s Tibo (@thsottiaux) posted that usage limits are doubled for all Pro 100/200/500 tiers and Business Premium, effective today, and that the planned Pro 200 usage-limit reduction will not happen. OpenAI’s help article (captured Oct 30, 2026, 8:20 AM CT) is narrower: Pro 100 and Pro 500 doubled; on Pro 200 only the lower allowance doubled; Plus is unchanged.¹source
- 2026-11-03OpenAI’s official Codex pricing page (developers.openai.com/codex/pricing, which redirects to learn.chatgpt.com/docs/pricing) states that ChatGPT Work and Codex share usage, and that ChatGPT Work usage inside ChatGPT uses the same pricing, credits and usage limits as Codex.source
Fireworks AI 4
- 2026-09-01Serverless adaptive rate-limit ceilings now vary by model size tier.source
- 2026-10-02Adaptive rate-limit ceilings for Small serverless models increased.source
- 2026-10-07Adaptive rate-limit ceilings for Small serverless models doubled.source
- 2026-10-18Small rate-limit tier threshold raised from <600B to <800B total parameters; ceilings unchanged.source
Amazon Bedrock (AWS) 2
- 2026-09-29Per-model tokens-per-day quota replaced by a cross-model, per-account, per-Region daily token quota.source
- 2026-10-12Quotas page now states applied quotas may be lower than listed defaults based on account age, usage and payment history, and increases aren’t automatic.source
Anthropic (Claude) 2
- 2026-09-22A rate limit reset for subscription users, saved and used when chosen ³entrysource
- 2026-10-15Fast mode for Claude Opus 5.5 left research preview and received a default rate limit in Start, Build and Scale tiers.source
xAI (Grok) 2
- 2026-09-28Voice models brought under the spend-based rate limit tiers.source
- 2026-10-13Per-second request limit formula loosened from RPM / 60 to RPM / 48 with a minimum of 2.source
GitHub Copilot 1
- 2026-10-08Org owners can bill Copilot code reviews to the organization instead of consuming members’ Copilot quotas.source
Hugging Face 1
Together AI 1
1. One post, counted once as one limit change: We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today.
The same post says Pro 200 usage limits will not be reduced, reversing the lower allowance in OpenAI’s help article; that planned reduction never took effect and was never a changelog entry, so it is not counted separately. OpenAI’s help article (captured Oct 30, 2026, 8:20 AM CT) describes the same Oct 29 change in more detail, so it is not a separate event: Pro 100 and Pro 500 now include twice their previous Work and Codex allowance. On Pro 200, only the lower allowance doubled.
It also says The reduction announced in our September 29 email is canceled.
and Plus usage is unchanged.
(help article) Posted Oct 29, 2026, 11:56 AM CT; reposted by @OpenAI at 12:12 PM CT (source; our entry; Pro 200 guide).
2. The 2026-10-13 post says "by EOD PST"; this page uses UTC dates, so in Pacific time that reset falls on Oct 12, 2026.
3. The same 2026-09-22 Anthropic (Claude) announcement also raised five-hour usage limits: In addition to the price drop, we’re increasing five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans.
(source; our entry). It is one event, counted once, as a reset.
Gemini weekly limit
Google’s Gemini Apps help page says a weekly limit exists. Its size is not stated, and the page gives no date for when it started, so it is not counted above. Checked Oct 29, 2026, 3:12 PM CT.
Your limit refreshes every 5 hours until you reach your weekly limit.
Method
- Counts come from LimitWatch’s own log only. Limit resets are rows of our AI usage limit resets log dated on or after 2026-09-01. Limit changes are our changelog entries in the rate-limit and quota categories detected on or after 2026-09-01. Pricing and plan entries are not counted.
- Same-event rule: when a reset row and a changelog entry have the same date and the same source URL, they are one event and are counted once, as a reset. This applies to 1 entry: Anthropic (Claude) 2026-09-22.
- Each row has its own primary source: an official company page, or a post by a named company staff member (tagged staff post, with the account). 11 of 12 resets and 1 of 15 limit changes are staff posts. The account @thsottiaux is labelled "Tibo, OpenAI" because its public X profile gives the name
Tibo
and the bioCodex & ChatGPT @OpenAI
(checked Oct 29, 2026, 8:27 AM CT). - Dates are the post or page date in UTC. "As of" is the latest date in either dataset; "checked" is when we last fetched the Gemini page, in Central Time (CT).
- A count reflects what LimitWatch has recorded, not every change a provider has made. No third-party tallies are used.
Sources
- docs.anthropic.com /en/release-notes/api
- docs.aws.amazon.com /bedrock/latest/userguide/doc-history.html
- docs.aws.amazon.com /bedrock/latest/userguide/quotas.html
- docs.fireworks.ai /updates/changelog
- docs.together.ai /docs/changelog
- docs.x.ai /developers/rate-limits
- github.blog /changelog/2026-10-08-copilot-code-review-new-organization-billing-options-and-controls
- github.com /huggingface/hub-docs/commit/c708a9e9fd
- learn.chatgpt.com /docs/pricing
- platform.openai.com /docs/changelog
- support.google.com /gemini/answer/16275805?hl=en
- anthropic.com /claude-opus-5-5
- x.com /thsottiaux
- x.com /thsottiaux/status/2103637477760311522
- x.com /thsottiaux/status/2107913674593644711
- x.com /thsottiaux/status/2109045392684032299
- x.com /thsottiaux/status/2109812360596865224
- x.com /thsottiaux/status/2110595484180214059
- x.com /thsottiaux/status/2111492673890824348
- x.com /thsottiaux/status/2112432076767289678
- x.com /thsottiaux/status/2113162277688070445
- x.com /thsottiaux/status/2113492187770175603
- x.com /thsottiaux/status/2114200152701575656
- x.com /thsottiaux/status/2114588174249071102
- x.com /thsottiaux/status/2115850175868727621
See also: Claude vs Codex limits, side by side · AI usage limit resets log · Codex limits: what’s confirmed · Claude limits draining faster? What’s confirmed · What each AI coding plan gets you. Something wrong? Dispute it; corrections are dated and appended.