Why Is Claude So Slow? 6 Causes and the Settings That Fix Them

Why is Claude so slow? Usually for one of four reasons. You picked a slower model, the effort level is set high, the chat has grown long, or Anthropic has an incident. Anthropic's own documentation ranks Claude Fable 5.1 as its slowest current model and Claude Haiku 5.5 as its fastest. Anthropic says higher effort means answers "take longer." Check status.claude.com first. If it is green, the fix is in the model menu next to the send button.

Most guides to this problem still tell you to switch off extended thinking. On the current Claude models you can't. Anthropic's help center says thinking cannot be turned off on Haiku 5.5, Sonnet 5.5, Opus 5.5 or Fable 5.1. The setting that changes speed now is effort. This guide covers how to tell an outage from a settings problem, which model to pick for speed, why long chats slow down, and what to change in Claude Code.

The Key Takeaways

• Check status first: Claude's status page logged 50 incidents between July 27 and October 7, 2026. Only one update said responses were "slower than normal." If the page is green, the cause is almost always your settings.
• Model choice is the biggest lever: Anthropic rates Fable 5.1 "Slower", Opus 5.5 "Moderate", Sonnet 5.5 "Fast" and Haiku 5.5 "Fastest".
• Thinking can't be switched off: on the 5.5 models and Fable 5.1, lower the effort level instead. Low and Medium are the quick settings.
• Long chats cost time: near the context limit Claude summarizes earlier messages, which shows up as "organizing its thoughts." A new chat is faster.
• Claude Code has its own fix: /clear, /compact and fast mode, which Anthropic says makes Opus up to 2.5x faster at a higher price.

Why Is Claude So Slow Today? Check Whether It Is Claude or You

Claude is slow for everyone only when Anthropic has an open incident, and that is the first thing to rule out. Open status.claude.com, Anthropic's official status page. It lists claude.ai, the Claude API, Claude Code and Claude Cowork as separate components, so you can see whether the app you are using is affected.

The status page rarely reports slowness as such. For the Fello AI status-page check, we pulled every incident from the page's public feed for the period from July 27 to October 7, 2026. There were 50 incidents. 14 were titled "degraded performance" or "elevated latency," and 13 of those listed claude.ai. When we read the updates, most of the 14 described error spikes, not slow answers. Only one update, on September 10, said some users were seeing "slower than normal response times," and that one concerned the Claude API for traffic entering through the US Midwest. The 14 incidents lasted about 68 minutes at the median.

So a green status page plus a slow Claude points at your own setup: the model, the effort level or the chat. The rest of this guide covers those. If Claude is returning errors or not answering at all, that is an outage problem. Our guide to what to do when Claude is down or not responding covers it step by step.

Which Claude Model Is Fastest?

Claude Haiku 5.5 is the fastest current Claude model, and Claude Fable 5.1 is the slowest. That ranking comes from Anthropic's own model comparison, which gives each model a "Comparative latency" label. Anthropic adds a caveat under the table: "Actual latency depends on prompt length, output length, and thinking effort." The label tells you the starting point. Your settings decide the rest.

Plan availability comes from Anthropic's pricing page. The Free plan includes Sonnet and Haiku, Pro adds Opus, and Fable sits behind usage credits on Pro and inside a share of the weekly limit on Max. A Free user who finds Claude slow is almost always on Sonnet 5.5, and switching that chat to Haiku 5.5 is the quickest single fix. Our guide to Claude Haiku 5.5 covers what the model is good and bad at.

Many people pick Opus or Fable because it sounds like the best model, then wait. Anthropic's own Claude Academy guide warns against using Opus on a task Haiku could handle. It says that choice "costs you tokens for no gain and slows you down in the process." Its help center adds that simple questions, basic information requests and general writing "don't need extra effort or thinking." Between the two heavy models, Anthropic labels Opus 5.5 "Moderate" and Fable 5.1 "Slower", and our Claude Fable vs Opus comparison covers when Fable is worth the wait.

Why Is Claude So Slow on Simple Questions? The Effort Setting

Claude is slow on simple questions when the effort level is higher than the question needs. Effort controls how much Claude thinks before it answers. Anthropic's help center puts the trade-off plainly. "Higher effort means more thorough responses, but they take longer and use more tokens, so you'll reach your usage limits faster." Anthropic's developer documentation calls effort "the main control for how much the model thinks, and with it quality, latency, and cost."

You Can't Turn Thinking Off Any More

On the current Claude models, the old "turn off extended thinking" fix no longer exists. Anthropic's help center is explicit. "Thinking cannot be turned off in Claude when using Claude Haiku 5.5, Claude Sonnet 5.5, Claude Opus 5.5, Claude Fable 5.1, or Claude Opus 5." On Opus 5.5 and Fable 5.1, thinking stays on at every effort level. When Claude is thinking, a "Thinking" indicator with a timer shows how long it has been working.

Anthropic knows long thinking frustrates people. In an April 2026 postmortem on Claude Code, Anthropic wrote that its high effort setting had caused latency long "enough to make the UI appear frozen" for some users. Anthropic briefly lowered the default to medium, then reverted after users said they preferred the smarter default. The same post states the trade-off directly: "In general, the longer the model thinks, the better the output."

How to Lower the Effort Level

Changing effort takes three clicks. Click the model name next to the send button, click "Effort," and choose a level. The change applies from Claude's next response, so you can switch in the middle of a chat. Each model marks its recommended level as "Default."

For everyday chat, Low or Medium is the right place to start. Raise effort only when an answer comes back shallow. Our breakdown of Claude Sonnet 5.5 and its effort setting shows how much the levels change on a single model.

Why Claude Slows Down in Long Chats

Claude gets slower in long chats because every new answer has to take in everything that came before it. Prompt length is one of the factors Anthropic says sets latency, and in a chat the prompt includes the whole history. Claude manages long chats automatically on paid plans with code execution turned on. Anthropic says that when a conversation "approaches the context window limit, Claude summarizes earlier messages to make room for new content." Anthropic adds that you may notice Claude "organizing its thoughts" during long conversations. That message is the summarizing step at work, and it is extra work before Claude answers.

The context window on Haiku 5.5, Sonnet 5.5, Opus 5.5 and Fable 5.1 is 1 million tokens on paid plans, so the summarizing usually starts only in very long chats. Big uploads fill it much faster than typing does. Tools and connectors add to it too: Anthropic calls them "token-intensive" and suggests being mindful of how many you have active.

Two habits help. Start a new chat when you change topic, and paste a short summary of the old one if Claude needs the background. For a large set of reference files, use a Project instead of uploading everything into one chat. Anthropic says Projects use retrieval, so Claude loads only the relevant parts into the conversation. Our guide to setting up Claude Projects walks through it.

Other Reasons Claude Feels Slow

Three smaller causes explain most of the slowness left after model, effort and chat length. They are answer length, an automatic model switch, and busy periods on Anthropic's side.

Long Answers Take Longer to Write

Claude often writes long, structured answers, and a longer answer takes longer to finish. Anthropic lists output length as one of the things that set actual latency. Anthropic's model documentation also says that if you prefer shorter responses, you should ask for them in your prompt. A line such as Answer in three sentences or Just the command, no explanation cuts the wait as well as the reading.

When Claude Switches Models Mid-Chat

Claude Sonnet 5.5 runs safety checks on every request, and a narrow set of flagged requests falls back to Claude Sonnet 5. Anthropic says the fallback mainly covers higher-risk cybersecurity work and some frontier AI development requests. When it happens, Anthropic says, "Claude re-runs it on Sonnet 5 in the same conversation," with a notice that the model switched. That re-run is a second pass at your request, so the answer arrives later than usual. The picker then stays on Sonnet 5 for the rest of the chat until you switch back.

Busy Periods and Your Plan

Anthropic acknowledges high-traffic periods by selling priority during them. Its pricing page lists "Priority access at high traffic times" for Max, Team and Enterprise, but not for Free or Pro. Anthropic's Pro help article, last updated in September 2026, still lists "priority access to Claude during high-traffic periods" for Pro, so the two pages disagree. Neither page promises faster answers. Our comparison of Claude Pro vs Max covers what else the plans change.

If none of this applies, run the basics. Update the Claude desktop app or reload the browser tab, try a different network, and test the same prompt on claude.ai in a browser. If the browser is quick and the app is slow, the app is the problem.

Why Is Claude Code So Slow?

Claude Code slows down as a session grows, because it sends the whole session with every request. Anthropic's Claude Code documentation says it "sends your full conversation with every request," and each round of tool use sends another request with the results. A session that has been open all day carries all of that into every new prompt, and Anthropic lists prompt length among the factors that set latency.

Four commands fix most slow sessions:

1. /clear starts a fresh session. Anthropic recommends it "when switching to unrelated work."
2. /compact summarizes the session to free space. You can tell it what to keep, for example /compact Focus on code samples and API usage.
3. /model switches to a faster model, such as Sonnet 5.5 or Haiku 5.5, for routine edits.
4. /fast turns on fast mode. Anthropic says it makes Claude Opus "up to 2.5x faster at a higher cost per token."

Fast mode has limits. It works only on Opus 5.5, Opus 5 and Opus 4.8. It costs $8 per million input tokens and $40 per million output tokens on Opus 5.5, double the standard rate. On Pro, Max, Team and Enterprise subscriptions it draws on usage credits, not on your plan's included usage.

If Claude Code feels both slower and less capable at once, check the effort level before anything else. The April 2026 postmortem traced a month of complaints to an effort default change and two other product changes, not to the model. Anthropic wrote: "We never intentionally degrade our models."

Is Claude Slower Than ChatGPT?

Claude often feels slower than ChatGPT because the current Claude models always think before they answer, and you can't switch that off. ChatGPT gives you a lighter setting to pick. OpenAI's release notes for GPT-6 describe settings running "from Instant through Extra High reasoning." On Claude, the closest match to ChatGPT Instant is Haiku 5.5 at Low effort.

We have not found an independent, current speed test comparing the two apps, so we don't quote one. The fair comparison is setting against setting: Claude Haiku 5.5 at Low against ChatGPT Instant, or Opus 5.5 at High against ChatGPT's higher reasoning levels. ChatGPT has its own slowdowns, covered in our guide to why ChatGPT is slow. For quality rather than speed, see our full Claude vs ChatGPT comparison.

How to Make Claude Faster: The Checklist

Work down this list in order. Each step takes under a minute, and the first three fix most slow chats.

1. Check status.claude.com. If claude.ai shows an incident, wait it out or use another model.
2. Switch model. Haiku 5.5 for quick tasks, Sonnet 5.5 for everyday work. Keep Opus and Fable for hard problems; our Opus vs Sonnet guide shows which tasks need Opus.
3. Lower the effort. Model menu, then Effort, then Low or Medium.
4. Start a new chat when you change topic, especially after Claude starts "organizing its thoughts."
5. Ask for shorter answers in the prompt.
6. Trim the context. Move reference files into a Project and turn off connectors you are not using.
7. In Claude Code, run /clear or /compact, and try /fast on Opus if speed matters more than cost.

Higher effort also uses your allowance faster, so these steps help with limits too. Our guide to Claude usage limits explains the five-hour and weekly meters.

Get Faster Answers With Fello AI

When Claude is slow and you need the answer now, the quickest fix is often a different model, not a different setting. Fello AI puts Claude, GPT, Gemini, Grok, Perplexity, DeepSeek, Kimi, GLM and Qwen in one app on Mac, iPhone and iPad. If Claude is stuck thinking or Anthropic has an open incident, you pick another model and keep working in the same app.

That also changes how you use Claude itself. Keep Claude for the long documents and careful writing it is good at, and send quick lookups, rewrites and one-line questions to whichever model answers fastest that day. You don't need a second subscription to do it: Fello AI has a free tier, and the full plan costs $9.99 a month or $79.99 a year.

Fello AI is rated 4.7 stars from more than 27,000 reviews on the App Store, and it can also search the web, read PDFs and Word files, and build PowerPoint decks and Excel sheets when the task needs more than a chat answer.

The Bottom Line

When Claude is slow and the status page is green, it is almost never Claude being broken. It is Claude doing what it was asked: thinking at a high effort level, on a heavy model, over a long chat. The old advice to switch thinking off is gone. The new fix is simple: drop to Haiku 5.5 or Sonnet 5.5, set effort to Low or Medium, and start fresh chats often. Save Opus, Fable and Max effort for the problems that need them, and accept the wait when you do.

Frequently Asked Questions