OpenAI is doubling usage limits for several paid ChatGPT plans and adding credits for its programming interface to eligible subscriptions. The changes mainly affect people and businesses that regularly use Artificial Intelligence (AI) for office tasks, coding, or automated workflows. At the same time, new reporting features from Anthropic show why larger allowances reveal little about the business value that spending produces.
What the new ChatGPT limits provide
According to the report on the updated ChatGPT allowances, OpenAI has doubled the Work and Codex limits for its Pro 100 and Pro 500 tiers. For Pro 200, the company doubled the lower allowance and reversed a planned reduction after receiving user feedback. The Plus plan is unchanged.
New subscriptions already include the revised allowances, while existing customers are supposed to receive them automatically. OpenAI also acknowledges that twice the allowance does not necessarily mean twice as many completed tasks. Consumption varies with the model, the job, and the selected settings.
In practical terms, summarizing a short document does not use the same capacity as a longer workflow involving several steps. Codex, OpenAI’s tool for programming tasks, can also consume different amounts depending on the assignment. A limit is therefore better understood as a usage boundary than as a dependable number of finished jobs.
OpenAI also says more efficient models should increase the amount of useful work available within those boundaries. That is a provider claim and has not been independently verified. The changes make the plans more generous, but they do not solve the underlying problem that users cannot always predict consumption before starting a task.
How subscriptions and API credits work together
In addition to chat limits, customers on certain OpenAI plans now receive credits for the Application Programming Interface (API). An API connects an AI model to another application instead of limiting its use to a chat window. According to Simon Willison’s summary of the API credit benefit, subscribers on the $100, $200, or $500 plans receive a corresponding amount of credit that can be assigned to one API organization.
Anthropic had previously announced a similar arrangement for its Max and Team subscriptions. Subscribers at the $100, $200, and $500 levels receive the same respective amounts in API credits, which expire according to the source. The available report does not specify an exact validity period for OpenAI’s credits, so Anthropic’s expiration rule should not automatically be assumed to apply to them.
The distinction between a subscription and API use still matters. A subscription covers activity in the included products and within their particular limits. API credits can instead support custom applications or public demonstrations, which Willison identifies as one concrete use case. Someone who works only in the chat interface may get little value from the additional credit, while a business could use it to partly fund an existing automated workflow.
OpenAI’s new always-on AI agents, called Dots, are also counted separately from the regular allowance. An AI agent is a system that carries out a task independently across multiple steps. Separate counters may prevent Dot activity from immediately consuming the standard limit, but they also make the overall view of usage and costs more complicated.
Why companies need closer measurement
Anthropic is addressing that complexity with new reports for enterprise customers. According to TechCrunch’s coverage of Anthropic’s spending tools, managers can review what a team produced during a selected period, how much it cost, and where its workflows failed. The reports combine internal analytics with employee surveys and are intended to recommend ways to improve results.
Anthropic says the underlying data will not be used for model training by default. This is a provider statement, not an independent privacy assessment. Companies handling internal work data should still decide what information is collected, who may access the reports, and how those reports may be interpreted.
Anthropic is also introducing pooled group budgets to help teams control their token spending. A token is a small unit of text that many AI services use to calculate consumption and cost. Shared budgets can improve visibility, but they do not by themselves establish whether an expensive workflow makes financial sense.
The figures cited by TechCrunch illustrate the pressure to improve oversight. In an analysis by software tooling company Jellyfish covering 276,000 customers, Claude Code and other tools based on Anthropic models accounted for nearly one-third of proposed code changes submitted for review. Spending on those tools had increased tenfold since the beginning of the year and was equivalent to about 2 percent of total engineering costs. TechCrunch also reports that most Fortune 500 companies are not yet seeing major gains from their AI deployments.
Swiss trade publication Netzwoche frames the development as a new phase of Financial Operations Management (FinOps), meaning the shared management and optimization of cloud costs. According to the State of FinOps Report 2026 cited in the article, 98 percent of FinOps teams now manage AI spending, compared with 31 percent two years earlier. The figures suggest that AI costs have moved from a specialist concern into routine budget management.
Pros and Cons of higher allowances
Pros:
- More headroom – Doubled allowances reduce the chance of hitting a plan limit early during demanding tasks.
- API credits – Eligible subscriptions can fund custom applications and automated workflows alongside chat use.
- Clearer attribution – Pooled budgets and reports can show which team creates costs and what results follow.
- Broader use – OpenAI says nearly one in five desktop Codex users opens features for documents, presentations, or spreadsheets rather than using it only for coding.
Cons:
- Unpredictable consumption – The model, input length, context, and settings can substantially change the cost of an individual task.
- Separate counters – Chat limits, API credits, and agent allowances do not produce one unified view.
- Uncertain return – More usage does not prove higher productivity or a positive return on investment, meaning measurable value relative to cost.
- New attack surface – Publicly accessible AI applications can be flooded with requests specifically to drive up expenses.
What this means for your cost control
For beginners, the most useful first step is to separate the main types of cost. Record the monthly subscription price, included API credit, extra API spending, and the primary concrete use case. If someone uses Codex to work on documents, presentations, or spreadsheets, for example, the number of requests should not be the only metric; the process should also produce a usable work result.
You should also check whether credits expire and whether they can be applied to the intended account or API organization. A larger credit is not a discount if you have no need for it. The same applies to a higher chat limit that you never approach during normal work.
Advanced teams can compare usage and outcomes for each workflow. Relevant measures include token consumption, model choice, failed attempts, and a business metric such as the amount of reviewed work. The FinOps analysis published by Netzwoche also recommends clear ownership, appropriate performance indicators, deliberate model selection, and caching for repeated requests.
Companies operating public AI services should set spending caps, request-rate limits, and automated alerts for unusual consumption patterns. Netzwoche describes deliberate cost attacks as denial of wallet: attackers flood an application with requests until its bill rises. A generous API credit may soften the effect temporarily, but it cannot replace technical and organizational controls.
Availability is an additional consideration for users in Switzerland. According to the report from OpenAI’s Berlin DevDay, personal Pro accounts in the European Economic Area, Switzerland, and the United Kingdom were excluded from Dots, while certain enterprise accounts could already receive access. OpenAI indicated that a broader European launch could follow within weeks after discussions with regulators, but it did not provide a separately confirmed date for Swiss personal accounts.
Higher limits and API credits make paid AI subscriptions more flexible, particularly when chat, automated workflows, and teamwork are used in parallel. Their financial value remains harder to measure than the number of available tokens or requests. The unresolved risk is rapidly growing, fragmented, and sometimes attackable spending that may remain invisible until the bill arrives.
Sources
- OpenAI verdoppelt ChatGPT-Nutzungslimits für Pro-Tarife und Business Premium – The Decoder, 2026-10-29
- API credits benefit for eligible ChatGPT subscriptions – Simon Willison’s Weblog, 2026-10-29
- Anthropic just launched tools to show companies what their AI spending is buying – TechCrunch, 2026-10-28
- DevDay in Berlin: OpenAI stellt Dots für Europa in Aussicht und treibt den Ausbau zur Plattform voran – The Decoder, 2026-10-28
- FinOps 2.0: Wie KI die Cloud-Kostenkontrolle herausfordert – Netzwoche, 2026-10-29
