Hit your Opus limit in Claude Code? Switch models and keep going

If Claude Code says a specific model's limit is used up, such as "You've hit your Opus limit · resets 3:45pm", switch to a model outside that family with /model and keep working. That works because the limit applies only to that model family. It does not work for the plain session or weekly message, which are shared across every model.

Two kinds of limit message

Claude Code's error reference lists these examples:

You've hit your session limit · resets 3:45pm
You've hit your weekly limit · resets Mon 12:00am
You've hit your Opus limit · resets 3:45pm
You've hit your Sonnet limit · resets 3:45pm

The first two are shared limits. The docs say "The session and weekly limits are shared across all models, so switching models doesn't restore access." The last two are model limits, and the docs say they "each apply only to requests to that model family, so switching to a model outside the family with /model keeps you working."

So the rule is simple: read the message. If it names a model, switch. If it says session or weekly, switching is pointless; see session limit options or weekly limit options.

What Anthropic documents today

There is a wrinkle. The Claude Code error page still lists the Opus and Sonnet limit messages, but Anthropic's Help Center describes only one separate per-model weekly limit as of November 2026: the one for Fable models. Anthropic's pages do not say whether any plan still enforces a separate Opus or Sonnet weekly limit today. The safe reading is: some plans have a separate weekly limit for one model, and if Claude Code says a model's limit is used up, another model keeps you working.

Some history, for context:

• Weekly limits took effect on August 28, 2025, with one overall limit and one for Opus 4 (TechCrunch, quoting Anthropic).
• On November 24, 2025, Anthropic said that for users with access to Opus 4.5, "we've removed Opus-specific caps" (Anthropic).
• After that, users asked on GitHub whether Sonnet now had a limit of its own (issue #12487), and a Max user described a separate weekly "Sonnet only" bucket (#55663). Those are user reports, not official behaviour.

If you read older advice that says "the separate weekly limit is for Opus", treat it as out of date.

The Fable limit

Max and Ultra plans, and premium seats on Team and on older seat-based Enterprise contracts, have a separate weekly limit for Fable models. Anthropic's wording: "You can use up to 50% of your weekly usage limits on Fable models at no extra cost. They draw from your plan's regular weekly usage limits and use them faster than other Claude models" (Fable models on your plan).

That 50% is a share, not bonus room. Anthropic says "you can never use more than your weekly limit", because your other models draw from the same pool. When you reach the Fable limit, the Help Center says you can "keep using Fable models with usage credits, or switch to another Claude model to keep working within your plan's usage limits."

On Pro, Team Standard seats and Standard seats on seat-based Enterprise, Fable models are normally not included in plan limits at all and run on pay-as-you-go usage credits from the start. The current exception is the Fable 5.2 launch offer: through November 16, 2026 at 11:59 pm PT, Pro and Team Standard seats have all Fable models included, up to 50% of weekly limits (Fable 5.2 launch offers).

When Fable would bill to usage credits, the /model picker shows "Requires usage credits" on the Fable row, and in an interactive session Claude Code can ask whether to continue on Fable using usage credits or switch to your default model. In -p runs it never asks and bills without asking (model configuration). More in the Fable weekly limit.

How to switch

Inside Claude Code:

/model

The picker saves your choice as the default for new sessions; press s in the picker to switch for this session only. From a shell, claude --model <alias or name> sets the model for one session. As of November 2026, on the Anthropic API the aliases resolve like this per the model configuration docs: opus is Opus 5.5, sonnet is Sonnet 5.5, haiku is Haiku 5.5 and fable is Fable 5.2. More options are in switching models in Claude Code.

The catch: the cache

Each model has its own prompt cache. Anthropic's error page warns that after switching models, "the next request re-reads the whole conversation with no cache hits". In a long conversation, that first request after the switch is a large one, and it counts against your limits like any other. If your context is big and you are near the shared session or weekly limit, a switch mid-conversation is an expensive step. If the next piece of work is a new task anyway, clearing with /clear first is cheaper than carrying a huge context across models; Anthropic notes that /clear "costs nothing".

Switching down also tends to use less

Anthropic says "Opus costs several times more per turn than Sonnet, and Sonnet more than Haiku" and that "Sonnet handles most coding tasks well and costs less than Opus." It publishes no exact multiplier in terms of plan limits, so treat that as direction, not a number. A common pattern is opusplan: Opus in plan mode and Sonnet for execution.

What about automatic continue

Claude Code's automatic continue, which waits in the session for a limit to lift, does not start on its own for an Opus or Sonnet limit while you run a model outside that family (interactive mode docs). The docs give the reason: your next turn may not hit that limit. You are not blocked, so there is nothing to wait for. That exception does not cover opusplan and other settings that run plan mode on the limited family, because their plan-mode turns would hit the limit again.

Keeping an eye on the per-model bar

If your plan has a per-model weekly limit, it is easy to miss until it blocks you. Headroom, a free Mac menu bar app, shows each Claude Code account's session and week percentages and, when the account has a per-model weekly limit, a meter labelled with the model name. When that model is used up while others still work, the row reads like "Fable back in 6d 3h" with a half-red dot. It reads the sign-in Claude Code already saved on your Mac and sends no prompts. It is independent and not made or endorsed by Anthropic. Get it at headroom.everydaymactools.com.