Claude · Pro · new quota2 models · shared allowance
$20 / moModelDial subscription data
Claude Pro vs Max 5× vs Max 20× usage limits
Plan pricing, allowance comparisons and usage estimates
3 plans
Models share the allowance; values are not additive. Tokens include cache.
Claude
3 plansSwipe to compare →| Weekly usage100M tokens · USD | Pro · new quota $20 / mo | Max 5x · new quota $100 / mo | Max 20x · new quota $200 / mo |
|---|---|---|---|
| Sonnet 5.5 | ≈9.69API equiv. ≈ $286 | ≈46.15API equiv. ≈ $1,366 | ≈97.62API equiv. ≈ $2,891 |
| Opus 5.5 | ≈6.69API equiv. ≈ $272 | ≈32.31API equiv. ≈ $1,321 | ≈66API equiv. ≈ $2,706 |
| Fable 5.1 | No sample | ≈3.69API equiv. ≈ $294 | ≈7.38API equiv. ≈ $573 |
| Sources | Sources & assumptionsSonnet 5.5 2026-10-05 · SemiAnalysis · agentic workload Reported monthly value; week normalized as month × 12 / 52 Original source workload; not converted to ModelDial tasks Opus 5.5 2026-10-05 · SemiAnalysis · agentic workload Reported monthly value; week normalized as month × 12 / 52 Original source workload; not converted to ModelDial tasks | Sources & assumptionsSonnet 5.5 2026-10-05 · SemiAnalysis · agentic workload Reported monthly value; week normalized as month × 12 / 52 Original source workload; not converted to ModelDial tasks Opus 5.5 2026-10-05 · SemiAnalysis · agentic workload Reported monthly value; week normalized as month × 12 / 52 Original source workload; not converted to ModelDial tasks Fable 5.1 2026-10-05 · SemiAnalysis · agentic workload; Fable is capped at 50% of the shared allowance, not the full plan Reported monthly value; week normalized as month × 12 / 52 Original source workload; not converted to ModelDial tasks | Sources & assumptionsSonnet 5.5 2026-10-05 · SemiAnalysis · agentic workload Reported monthly value; week normalized as month × 12 / 52 Original source workload; not converted to ModelDial tasks Opus 5.5 2026-10-05 · SemiAnalysis · agentic workload Reported monthly value; week normalized as month × 12 / 52 Original source workload; not converted to ModelDial tasks Fable 5.1 2026-10-05 · SemiAnalysis · agentic workload; Fable is capped at 50% of the shared allowance, not the full plan Reported monthly value; week normalized as month × 12 / 52 Original source workload; not converted to ModelDial tasks |
Versions & estimation notes
OpenAI: only Pro $200 has new and previous quotas. Plus and Pro $100 are unchanged; Pro $500 is a new tier. The source reports pre-cut subscribers retain old limits until October 29.
Claude: estimates use an agentic workload. Fable is limited to 50% of the allowance. Historical samples are retained for comparison and cannot be added to new quotas.
GLM: ranges use a 95% cache-hit rate, all peak to all off-peak. Average months use 52/12 and exclude promotions ending October 7. Missing API equivalents remain unavailable.
Quota calculator
Selected planClaude · Pro · new quota
Sonnet 5.5 · Weekly
%
Pro · new quota · Sonnet 5.5 · Weekly
Remaining tokens≈ 9.69 × 100M tokens
API equivalent≈ $286.38
Monthly allowance comparison
Filter plans by monthly budget and compare model allowances and API equivalents.
Similar price: monthly fee within ±5% of budget. Filters use original billing currencies without conversion; only plans with allowance samples are shown, with legacy samples labelled.
Similar-price capacity · $20
1 plans · Average month
Estimated API equivalent / month (USD) · Shared allowances, not additive
Metric definitions
API equivalent estimates the API cost of sampled monthly usage in USD; it is not an available balance. Model pricing, caching and workloads differ, so these values do not measure model quality. Models share each plan’s allowance; values are not additive.
About these plans
Which Claude model is compared here?
Current references separate Sonnet 5.5, Opus 5.5 and Fable 5.1; Fable is capped at 50% of the allowance. Earlier Opus samples remain as historical references. The monthly fee belongs to the subscription; tokens and API equivalents describe the corresponding usage samples. Tasks, cache mix and observation dates still affect the results. These are not fixed limits for every account.
Why do token totals and API equivalents differ so much?
Usage includes cache reads. Cached input, fresh input and output have different rates. Reusing long cached contexts can produce large token totals. Read tokens and dollars together, and expand the sources to inspect the sample mix.
Does reaching the five-hour limit exhaust the week?
The short window and weekly window are separate limits. An exhausted session does not mean the week is exhausted; an exhausted weekly allowance is not restored by waiting for the next short window alone. The table separates these periods and leaves missing records blank.