Opus priority access used up: what it means
Your weekly Opus allowance is spent. Sonnet still works. When Opus resets, whether credits or a Max upgrade bring it back early, and what to do in Claude Code.
What the Opus priority access notice means
Opus priority access used for the week means one thing: the separate weekly Opus allowance on your plan is spent. Nothing else has changed. Sonnet and Haiku keep working, your conversation is still there, and the general weekly limit that covers every model is a different counter that you have not necessarily reached. When the Opus counter resets, Opus comes back on its own.
The exact wording differs by surface, and Anthropic adjusts it from time to time. In the Claude app the notice says that your priority access to Opus is used for the week and that Sonnet stays available until your weekly limit resets. In Claude Code, as of 19 September 2026, the string in the transcript is shorter:
You've hit your Opus limitTwo other strings look similar and mean something different. You've hit your session limit is the five-hour window, shared across every model, and switching model does not help. You've hit your weekly limit is the all-models weekly window, also shared, and switching model does not help there either. Only the Opus message (and its twin, You've hit your Sonnet limit) is about one model family, which is why a model switch clears it. Claude Code's cost documentation says it plainly: after the model-specific message, switching to a model outside that family with /model keeps you working. How the session window and the weekly window fit together is covered in how Claude's five-hour and weekly usage limits work, so this page stays on the Opus counter.
Anthropic's plan pages describe access to Claude on paid plans as priority access. That is the benefit the notice is named after. Opus is still on your plan and your account is not flagged. The meter behind that benefit is at 100 percent for this week, and that is all the notice reports.
Which plans show the Opus notice
As of 19 September 2026, the help center says the Settings > Usage page exists for Pro, Max, Team and seat-based Enterprise plans, and that its weekly section shows when the limit resets "for Opus only and all other models". The Opus-only bar is the counter behind this notice. If your Usage page shows an Opus-only bar, your plan carries a separate Opus allowance and you can see this notice. If it shows a single weekly bar, your plan meters Opus inside the general weekly limit, and the message you will meet is the weekly one, not this one.
Anthropic has changed the shape of these counters more than once since late 2025, and the help center does not publish the size of the Opus allowance in hours or tokens. Any figure you read elsewhere is an estimate from someone's own usage. Treat the bars on your own Usage page as the only source of truth for your account. What each plan includes, and how the Max tiers differ, is in which models Pro and Max actually give you.
When does the Opus counter reset, and where is the timer
The Opus counter is a weekly counter. The help center states that the weekly limit "resets at a fixed time each week that is assigned to your account", and that you can see the next reset time in Settings > Usage. The Opus-only bar has its own reset time next to it. Read that time, and do not assume it matches the time on the all-models bar.
The timer is shown in two places:
- The notice itself. The limit message in Claude Code names the reset time, so the first place to look is the line that stopped you.
- The usage bars. On claude.ai, open Settings > Usage from your initials in the lower left. Inside Claude Code, run
/usage, which on a Pro, Max, Team or Enterprise plan draws the same plan bars plus a breakdown of what consumed them.
There is no way to reset the counter early from your side. It is a clock. Credits let you pay past it, but nothing moves the clock. Whether any usage limit can be reset on request has the longer answer, and it is no.
Do usage credits or a Max upgrade restore Opus right now?
Two different questions, with two different answers.
Usage credits do not reset the Opus counter. They let you pay past it. Usage credits are available on Pro and on both Max tiers. They are billed at standard API rates and appear as extra charges on your bill. The help center is explicit that credits do not change reset timing: your included limit still resets on its own schedule, and credits only change what happens in the meantime. It makes no exception for the Opus counter, so with credits switched on and a balance available, the Opus limit stops blocking you and each Opus request is metered against your balance instead. Opus is the expensive model to run on credits, so check the per-token price before you leave a long session on it. How usage credits are priced and capped walks through the spend limit and what a typical day costs.
One side effect is easy to miss. While you draw on credits, the prompt cache lifetime in Claude Code drops from one hour to five minutes. A long conversation resumed after a coffee break re-reads its whole history at the uncached rate, and that is billed. Keep sessions short on credits, or /clear between tasks.
A Max upgrade takes effect immediately, but the help center makes no promise about the counters. It says a Pro plan "will be upgraded to Max immediately, with billing adjusted on a prorated basis". It does not say that your current usage bars reset on upgrade. Check your Usage page straight after paying and read the Opus-only bar again. If the bars did not move, upgraded but the limits did not change covers the sign-out and sign-in steps and the cases that need support. Before you pay, what to check before upgrading to Max is worth five minutes, because if you filled the Opus bar on Pro in one heavy afternoon, a bigger bar changes the day you fill it rather than whether you do.
What to do inside Claude Code
As of the September 2026 documentation, Claude Code shows the message, stops the turn, and offers a menu. The docs describe no automatic switch to Sonnet. Older builds from mid-2025 did switch silently, with a transcript line reading Claude Opus 4 limit reached, now using Sonnet 4, which is one more reason to check your version below. What you do next depends on how far away the reset is.
Switch model explicitly. The direct fix is one command:
/model sonnetSonnet is outside the Opus family, so the next request goes through. During a usage-limit wait, switching with /model makes Claude Code re-check whether usage is available and continue right away. If Claude Code shows a cache warning first, confirm it. A prompt cache belongs to one model, so the first Sonnet request re-reads the conversation at the uncached rate, and that draws from the Sonnet and all-models counters. On a session that has been open for hours, that one request is not small.
Or split planning and execution. opusplan uses Opus while you are in plan mode and Sonnet for execution:
/model opusplanThis is the setting to leave on for the rest of the week once Opus is back, because it spends the Opus counter on the part of the work where model quality matters most. One caution from the docs: if you are on opusplan when the Opus limit hits, Claude Code does not use the early-continue path. It waits for the reset instead. Which tasks earn Opus and which run fine on Sonnet or Haiku is the subject of choosing between Opus, Sonnet and Haiku.
Or wait, and let Claude Code continue for you. On Claude Code v2.1.234 or later, when a usage limit stops a task, Claude Code can wait in the open session and pick the task up after the reset. While it waits, the bottom line reads:
Usage limit reached · continuing automatically at 3:45pm · esc to cancelTwo things stop it from starting that wait on its own. A reset more than 24 hours away, which a weekly counter often is. And an Opus or Sonnet limit while you are running a model outside that family, because your next turn may not hit the limit at all. In both cases Claude Code opens the usage-limit options menu once per reset window. Pick the row that starts Wait here, then continue automatically to start the wait anyway, or open the menu yourself:
/rate-limit-optionsPress Esc at an empty prompt, or Ctrl+C, to cancel a running wait. Claude Code confirms with a line starting Automatic continue cancelled. To turn the behaviour off for good, set autoContinueAtUsageLimit to false in your user settings, or switch off Continue automatically at usage limit in /config.
If you add credits or upgrade, do it from inside the session. During a wait, finishing /usage-credits, signing back in after /upgrade, or switching with /model makes Claude Code re-check whether usage is available and continue right away. It does not re-check after an upgrade or purchase you make in a browser on your own. If you paid in the browser, run /login again or restart Claude Code so it sees the new plan.
Check your version first, because the automatic wait and the options menu are recent:
claude --versionAnything older than 2.1.234 shows the limit message and stops, with no wait offered. Update, or switch model by hand.
Is in-flight Opus work lost?
No. The limit stops the next request, not the session. In Claude Code, the conversation, the edited files, the task list and the session itself all stay where they are. When the limit lifts, by reset or by a model switch, Claude Code sends a fixed prompt asking Claude to pick the task up where it stopped. It does not resend your last message, so if the interrupted turn was half way through a multi-step change, read the last few lines of the transcript before you continue and say what still needs doing.
Two details from the docs are worth knowing. If the wait ran while your computer slept for more than about 30 minutes, the line reads Your usage limit has reset · press enter to continue, and nothing moves until you press Enter. And the continued task still asks for permissions as usual, so an unattended session can stop on a permission prompt and sit there. If Claude Code hits the limit again on the continued task, it re-arms the wait at most twice in a row, then stops with a line starting Automatic continue stopped after repeated usage-limit hits.
In the Claude app, the chat stays open and the model picker at the bottom of the conversation still works. Choose Sonnet and send the next message in the same thread. The context Opus built is carried forward, because the conversation history is what gets sent, and the model that wrote it does not matter. The one thing you cannot do is make Opus finish the reply it was in the middle of. That reply is gone, and Sonnet writes the next one.
Reading the Usage page without guessing
The Usage page can show more than one weekly bar, and the notice you get depends on which one filled first. The all-models bar is the ceiling for everything. The Opus-only bar meters Opus inside that ceiling. In Claude Code there is a Sonnet-only counter too, which is why You've hit your Sonnet limit exists. On Max plans, Fable models draw from the same weekly limits, and the help center caps their included share at 50 percent of the weekly limit, after which you can keep using Fable on credits or switch to another model.
The practical reading rule: look at which bar is at 100 percent. Opus-only full and all-models not full is this page, and Sonnet is the answer. All-models full is the weekly limit, and the answers are waiting, credits, a bigger plan, or a lighter model mix next week. Both full at once means the same, and the Opus notice is the least of your problems. If you are filling the Opus bar every week on a pattern that will not change, compare the plan cost with paying per token in when the API costs less than the subscription, because a subscription that is capped for three days of every seven is not delivering the price you are paying for.
FAQ
Does the Opus priority access notice mean my whole account is limited?
No. It means the separate weekly Opus counter on your plan is at 100 percent. The all-models weekly limit and the five-hour session limit are different counters. Unless you also see You've hit your weekly limit or You've hit your session limit, Sonnet and Haiku keep working. Switch model and continue.
When does Opus come back after the notice?
At the reset time shown next to the Opus-only bar in Settings > Usage on claude.ai, or in /usage inside Claude Code. The help center says weekly limits reset at a fixed time each week assigned to your account. The Opus-only bar has its own time, so read that one rather than the all-models time. Nothing you do on your side moves it earlier.
Will buying usage credits give me Opus back immediately?
Credits do not reset the counter. They let you continue past it, billed at standard API rates, and the help center makes no exception for Opus. Turn them on in Settings > Usage, or with /usage-credits in Claude Code, and the next Opus request is metered against your balance. Expect Opus to be the most expensive model to run this way, and note that the prompt cache lifetime drops to five minutes while you are on credits.
What does Claude Code do when it hits the Opus limit mid-task?
It stops the turn, shows You've hit your Opus limit, and on v2.1.234 or later opens the usage-limit options menu. Run /model sonnet and the task continues on the next request. Or pick Wait here, then continue automatically and Claude Code resumes after the reset with a fixed prompt to pick the task up. Your conversation and edited files are not lost either way.
Does upgrading from Pro to Max reset my Opus counter?
The upgrade applies immediately and is prorated, but the help center does not promise a counter reset. Check Settings > Usage straight after paying and read the Opus-only bar. If the bars did not move, sign out and back in, then contact support if they still show the old plan.