Why is Claude so slow? Why it’s taking so long to respond, and how to speed it up

Claude is usually slow for one of four reasons: Anthropic’s servers are busy or having an incident, you are using a high effort level, you picked one of its heavier models (Fable 5.2 or Opus 5.5), or the chat has grown very long. Check status.claude.com first. If it’s clear, lower the effort level or switch to Sonnet 5.5 or Haiku 5.5, and start a new chat.

Why is Claude so slow today? Check the status page first

If Claude was fine yesterday and crawls today, the problem is probably on Anthropic’s side. Its status page lists live incidents for claude.ai, Claude Code, Claude Cowork and the API separately, and you can subscribe to updates there. Incidents are common: the page logged 17 in October 2026, and 10 of them were errors or slowdowns on particular models, such as Sonnet 5.5 on October 27 or Fable 5.1 on October 25. Most last minutes, though some, like the September 29 outage, take out sign-in and Claude Code for longer.

Because incidents often hit one model at a time, switching model in the menu next to the send button is often the quickest way to keep working. If the status page is all green and only you are affected, refresh the page, try the desktop or mobile app, and turn off a VPN or browser extensions, which Anthropic’s error troubleshooting page suggests for sign-in problems.

Claude taking so long to respond? Lower the effort level

The setting that most changes Claude’s speed is effort. Anthropic’s model and effort help page says: “Higher effort means more thorough responses, but they take longer and use more tokens.” You change it in the model menu next to the send button, and the new level applies from Claude’s next reply, even mid-chat.

• Low and Medium “work well for routine tasks and stretch your usage further”. Medium is the default in the Claude apps on Fable 5.2 and Sonnet 5.5.
• High, in Anthropic’s words, “offers the best overall balance of quality and speed”.
• Extra high and Max are for hard problems. Anthropic says to “choose Max effort and expect longer response times”.

If Claude seems stuck on “Thinking”, it probably is thinking. While it reasons you see a “Thinking” indicator with a timer, then an expandable Thinking section above the answer. On Fable 5.2, Opus 5.5, Sonnet 5.5, Haiku 5.5, Fable 5.1 and Opus 5, Anthropic says thinking “cannot be turned off”, so the thinking toggle disappears. To make these models think less, lower the effort level instead.

Which Claude model is fastest?

Haiku 5.5 is Claude’s fastest model and Sonnet 5.5 the next fastest. Anthropic’s models overview rates each one’s “comparative latency”:

From Anthropic’s models overview and the plan comparison on claude.com/pricing, checked November 11, 2026.

Anthropic adds that “actual latency depends on prompt length, output length, and thinking effort”, so a long chat on Haiku can still be slower than a short one on Opus. Each new model is also quicker than the one it replaced: Anthropic says Fable 5.2 writes up to 30% faster than Fable 5.1, and Opus 5.5 and Sonnet 5.5 more than 30% faster than their predecessors. For everyday questions, Sonnet 5.5 on Low or Medium is the quickest way to get a good answer.

Long chats, “organizing its thoughts” and big files

Every reply means Claude reads the whole conversation again, including files you uploaded, and the latency note above says prompt length counts. All current models have a 1 million token context window in the apps. When a chat gets close to it, Claude pauses to summarise earlier messages, and its usage and length limits page says you may see Claude “organizing its thoughts” while it does. That step needs code execution turned on, and Anthropic warns that longer chats which trigger it “consume more of your usage limit”.

The fix is to start a new chat when a topic changes, and to paste in a short summary rather than carrying on in one huge thread. The same page notes that “tools and connectors are token-intensive”, so turning off connectors and tools you aren’t using also lightens each request.

Some features are slow by design. Research runs on Opus 5.5 whichever model you have selected, and Anthropic’s help centre describes it making five or more tool calls over one to three minutes; when it launched, Anthropic said complex reports could take up to 45 minutes. Web search adds time too: Anthropic says search times “may vary based on query complexity”.

“Due to unexpected capacity constraints”: when Claude is busy

When too many people are using a model at once, Claude may slow down or show the message “Due to unexpected capacity constraints, Claude is unable to respond to your message. Please try again soon.” Anthropic says capacity constraints “are not outages”, that they “typically resolve as demand patterns shift throughout the day”, and that you should try again in a few minutes. Crucially, they “will not appear on our status page”, which is why Claude can feel sluggish while the page shows everything operational. Anthropic’s plan comparison lists “priority access at high traffic times” on the Max and Ultra plans.

Why is Claude Code so slow? Fast mode, effort and compacting

Claude Code has more speed controls than the chat apps:

• Fast mode. Type /fast to switch it on. Anthropic’s fast mode docs say it makes Fable 5.2 and Opus 5.5 “up to 2.5x faster” with the same quality, but not Sonnet or Haiku. On Pro, Max, Ultra, Team and Enterprise it is paid “via usage credits only”, at fast mode rates, even if you have plan usage left, so turn on usage credits first and switch it on at the start of a session, when it costs least.
• Effort. Run /effort and pick Low or Medium. Fable 5.2, Opus 5.5, Sonnet 5.5 and Haiku 5.5 default to Medium; if you raised it, lower it again for routine work. Thinking can’t be switched off on these models.
• “Opus is experiencing high load”. Claude Code retries busy errors itself before you see them. If it shows “Repeated 529 Overloaded errors”, Anthropic’s error guide says this “is not your usage limit”, and to use /model to switch model, “since capacity is tracked per model”.
• Stuck on compacting conversation. Compacting sends a separate request that summarises the whole session, so it takes longer the bigger the context, and longest when you resume an old session, because nothing is cached. Run /compact yourself at a natural break, or /clear for a fresh start, which costs nothing. Our guide to making Claude Code limits last covers more ways to keep the context small.
• High CPU, memory or freezes. Anthropic’s troubleshooting page suggests running /compact regularly, restarting between big tasks, adding large build folders to .gitignore, and starting with claude --safe-mode to see if a plugin, MCP server or hook is to blame. Above 2.5GB of memory, restart and run claude --continue. If it hangs, press Ctrl+C, then claude --resume. /doctor checks your setup.

Why is Claude slower than ChatGPT?

Often it isn’t a fair comparison. Claude’s current models always think before answering, and many people compare a heavy Claude model with ChatGPT’s Instant chat, its quick everyday option, which OpenAI even keeps off the meter in its new shared usage limits. Fable 5.2 on High effort against ChatGPT Instant is a long letter against a text message. For quick questions, compare Haiku 5.5 or Sonnet 5.5 on Low effort instead.

Claude slow in the US, UK, India, Australia or Europe

The causes and fixes are the same in every country where Claude is offered, including the US, Canada, the UK, Germany and the rest of the EU, Australia and India. There is one status page for everyone. Anthropic doesn’t publish its busiest hours, but because capacity limits ease “as demand patterns shift throughout the day”, trying again later can help wherever you are. A VPN can make things worse: besides adding a hop, using Claude from an unsupported country breaks its rules, as our guide to where Claude is available explains.

Quick fixes for a slow Claude

1. Check status.claude.com.
2. Switch model: Sonnet 5.5 or Haiku 5.5 for speed.
3. Lower the effort level to Low or Medium.
4. Start a new chat instead of continuing a long one.
5. Turn off tools and connectors you don’t need.
6. In Claude Code, try /fast, /compact at a break, or /clear.
7. Seeing a capacity error? Wait a few minutes and try again.

Last checked: November 11, 2026, against Anthropic’s status page, its help pages on effort and thinking settings, usage and length limits, error messages and Research, its models overview and pricing page, and the Claude Code fast mode, model configuration, errors and troubleshooting docs.

Frequently asked questions

Why is Claude so slow today?

Check status.claude.com first: Anthropic lists incidents there for claude.ai, Claude Code and the API, often affecting one model at a time. If it shows no incident, Claude may be busy, which Anthropic says does not appear on the status page, or your effort level, model or long chat is slowing it down.

How do I make Claude respond faster?

Pick Sonnet 5.5 or Haiku 5.5 in the model menu, set effort to Low or Medium, start a new chat for each topic and turn off tools and connectors you don’t need. In Claude Code, /fast makes Fable 5.2 and Opus 5.5 up to 2.5 times faster, paid from usage credits.

Can I turn off thinking in Claude?

Not on current models. Anthropic says thinking cannot be turned off on Fable 5.2, Opus 5.5, Sonnet 5.5, Haiku 5.5, Fable 5.1 or Opus 5. Choose a lower effort level to make them think less.

What is the fastest Claude model?

Haiku 5.5. Anthropic rates it “Fastest” in its models overview, Sonnet 5.5 “Fast”, and Opus 5.5 and Fable 5.2 “Moderate”. Haiku 5.5 and Sonnet 5.5 are on the Free plan as well as paid plans.

Why does Claude say it is “organizing its thoughts”?

Your chat is near its context limit, so Claude is summarising earlier messages to keep going. Anthropic calls this automatic context management. It uses more of your usage limit, so start a new chat if the conversation is very long.

Why is Claude Code stuck on compacting conversation?

Compacting sends a separate request that summarises the whole session, so it takes longer the more context there is, and longest after a break, when nothing is cached. Run /compact yourself at a natural break, or /clear if you don’t need the history.

What does “Due to unexpected capacity constraints” mean on Claude?

Too many people are using Claude at once. Anthropic says it is not an outage, usually clears as demand shifts during the day and won’t appear on the status page. Try again in a few minutes or switch model.

Does a 529 overloaded error use up my Claude limit?

No. Anthropic’s Claude Code error guide says a 529 is not your usage limit and doesn’t count against your quota. Capacity is tracked per model, so switching model with /model usually lets you keep working.

Is Claude slower than ChatGPT?

It can feel that way, because Claude’s current models always think before answering and people often compare a heavy model such as Fable 5.2 with ChatGPT’s quick Instant chat. Haiku 5.5 or Sonnet 5.5 on Low effort is the fairer comparison for quick questions.