ChatGPT Plus vs Pro vs Claude Pro: Usage Limits Explained (October 2026)
Published
Both plans cost $20 a month, so pick by the model you need. Then check the limit, because both companies change theirs often. Neither company publishes a fixed monthly message count. OpenAI gives a range of messages for each five-hour window on Plus. Anthropic gives only multiples of its own base plan.
Two things changed recently. Some Pro 200 subscribers lose usage on 30 October. Some GPT-6.1 Sol users also report that their allowance runs out faster since 16 October. We found no OpenAI explanation for that second change.
Pro 200 gets less usage from 30 October for subscribers who kept the old allowance
On 29 September, OpenAI added Pro 500 at $500 a month. It is the top tier. It is the only Pro plan with Ultrafast, a speed setting that uses more of your allowance (OpenAI help).
Pro 200 still costs $200 a month. New Pro 200 subscribers already get a lower allowance than the plan used to offer. People who had Pro 200 at any point from 22 September to 10 a.m. Pacific time on 29 September keep the old allowance until 29 October. After that, their allowance drops too (OpenAI help).
Press reports give the size of the cut. Pro 200 falls from 20 times to 10 times the Plus allowance, for Work and Codex. In chat, GPT-6 Pro messages fall from 200 to 100 a week (TNW). OpenAI’s help page does not print either multiple, so treat these two figures as reported, not confirmed.
OpenAI gives message ranges for every five hours, and no weekly number
On OpenAI’s Codex and Work pricing page, Plus gets these estimates of local messages for each five-hour window:
- GPT-6.1 Sol: 15 to 160 messages
- GPT-6 Sol: 15 to 150 messages
- GPT-6 Luna: 350 to 3,000 messages
- GPT-6 Astra: 5 to 45 messages, in Work and Codex only
OpenAI says these are estimates, not fixed limits. Local messages and cloud chats share one allowance. Pro plans have no five-hour limit on that page (OpenAI pricing).
For Chat, OpenAI’s help article gives no message count for Plus. Plus gets GPT-6 Sol at Medium and High thinking, but not Extra High (OpenAI help). So the answer to “what is the weekly limit for ChatGPT Plus?” is that OpenAI publishes no weekly number on the pages we read. It says weekly limits may also apply.
| Plan | Price a month | Usage compared with Plus | Usage per dollar (Plus = 1.0) |
|---|---|---|---|
| ChatGPT Plus | $20 | Base | 1.0 |
| Pro 100 | $100 | Not stated | Not stated |
| Pro 200, old allowance (grandfathered, until 29 October) | $200 | 20 times (reported) | 2.0 |
| Pro 200, lower allowance (new subscribers, and grandfathered from 30 October) | $200 | 10 times (reported) | 1.0 |
| Pro 500 | $500 | 25 times (reported) | 1.0 |
Usage per dollar is the multiple divided by the price, scaled so Plus equals 1.0. Each ratio rests on press figures, and each one is measured against Plus, so none of them tells you how many tasks you can finish.
After 30 October, paying more for Pro 200 buys no more usage per dollar than Plus. Pro 500 costs the same per dollar as Plus. You get a higher ceiling and Ultrafast, not a better rate.
Some GPT-6.1 Sol users report faster drain, and we found no OpenAI explanation
From 16 October, several Plus users wrote in an OpenAI community thread that Sol used up their allowance faster. One estimated about double. Another reported a 22% increase. A third estimated four to five times faster. The posters call these estimates, not controlled tests. The last reply we read is from 18 October. No OpenAI staff member replies in the thread (OpenAI community).
Three explanations fit the reports. OpenAI may count each task against more allowance. The total may have shrunk. Or users may have moved to cloud chats, which share the same allowance. We cannot tell which from outside. Until OpenAI says, treat this as a hypothesis.
Two documented settings can explain a jump on their own. OpenAI says usage depends on the model, the context, reasoning, tool use, retrieval, caching and speed mode. Fast mode uses 2.5 times the included allowance of Standard mode. Ultrafast uses 8 times for GPT-6.1 Sol and GPT-6 Astra (OpenAI help). Check your speed setting before you blame the model.
Claude gives multiples, not counts, and Max 20x gives twice Pro’s usage per dollar per session
Claude Pro costs $20 a month. Billed yearly, it costs $17 a month, or $200 up front (Anthropic pricing). Max costs $100 for 5 times Pro’s usage (Anthropic pricing), or $200 for 20 times (Anthropic help).
Anthropic says Pro gives at least 5 times the usage of Free in each five-hour session. Max gives 5 or 20 times the usage of Pro, also per session. Weekly limits apply on top, and Anthropic does not publish their size. Anthropic says there is no fixed message count. Your usage depends on how long and complex your conversations are, which model you pick, and which features you use. Web, desktop, mobile and Claude Code all draw from one pool (Anthropic pricing).
On the same scale, measured per five-hour session, Pro is 1.0 per dollar. Max 5x is also 1.0. Max 20x is 2.0, because it gives 20 times Pro’s usage for 10 times the price. This is our arithmetic on Anthropic’s numbers. It leaves out the weekly caps, which Anthropic does not publish.
Read this comparison only inside Claude. Anthropic counts per session, with weekly caps on top. OpenAI counts against Plus, and its Pro 200 multiple applies to Work and Codex. A 20 times on Claude and a 10 times on ChatGPT are different measures, so you cannot put them side by side.
Light and daily use fit a $20 plan; heavy coding needs $200 or more
This is our reading of the published numbers, not a test we ran.
- Light use, a few questions a week: OpenAI says Free and Go users get unlimited everyday text chats, on its lowest-cost model, subject to abuse safeguards (OpenAI help). A $20 plan is enough for more. Choose by the model you need.
- Daily use, research or writing most days: a $20 plan is still enough for most people. On Claude, if you hit session limits, Max 5x at $100 gives 5 times Pro’s usage per session.
- Heavy coding, hours a day in Codex or Claude Code: on Claude’s own per-session numbers, Max 20x at $200 gives the most usage per dollar of Claude’s plans. Weekly caps are not published, and heavy coders may hit them first. On ChatGPT, Pro 500 at $500 is the only tier with Ultrafast. Pro 200 gives no more usage per dollar than Plus once the lower allowance applies.
Five settings change how fast your limit runs out
- Turn off Fast and Ultrafast when you do not need speed. They use 2.5 and 8 times the included allowance (OpenAI help).
- Send simple questions to a lighter model. On OpenAI’s Codex and Work pricing page, Plus gets 350 to 3,000 GPT-6 Luna messages per five hours. GPT-6.1 Sol gets 15 to 160 (OpenAI pricing).
- Check the reset time before a big job. ChatGPT shows the reset time when it has one. OpenAI points to your usage dashboard for current limits (OpenAI help; OpenAI pricing).
- On Claude, pick the model that fits the task. The model you pick changes how fast the limit goes (Anthropic pricing).
- On Claude, turn off features you do not use. Features count toward usage, and Claude Code draws from the same pool as chat (Anthropic pricing).
LodeHQ’s daily AI newsletter covers new models, tools and research.
For the models themselves, read our comparison of Gemini 4 Argon, Claude Opus 5.5 and GPT-6 Astra.
How we checked
We read OpenAI’s Pro tiers help page and its pricing page on 20 October 2026. We read Anthropic’s pricing page on the same day. The Pro 200 and Pro 500 multiples come from press coverage, because OpenAI’s help text does not print them. The drain reports come from OpenAI’s community forum, read 20 October. We did not run either plan. Prices and limits change often. We will check them again on 31 October.