Try Kimi at kimi.com →

Kimi AI Not Working?
Common Problems and Fixes

Kimi won't load, says “server busy”, won't send a login code or stops making slides? Start with the six quick checks, then find your exact problem below, from the chat and apps to Kimi Code and the API.

25 problems solved API error codes Kimi Code fixes
kimi.com Summarise this report !Server is busyPlease try again later K ✓ Kimi is working again QUICK CHECKS Kimi status page No outage reported Internet connection Other sites load fine Browser cache & extensions Cleared, VPN off ! Plan limits & credits Credits left this month Try again in a new chat Kimi replies
QUICK

6 quick fixes to try first

These fix most problems in a couple of minutes. Work through them in order.

  1. 1

    Check if Kimi is down

    Look at a status tracker or search “Kimi down” on X. If many people report problems, wait; nothing on your side will fix an outage.

  2. 2

    Refresh and start a new chat

    Reload the page, then try again in a fresh chat. Very long chats can slow down or fail.

  3. 3

    Try another browser or the app

    If kimi.com works elsewhere, the problem is your browser: clear cookies for kimi.com and turn off ad blockers.

  4. 4

    Turn your VPN off (or on)

    Some VPNs, proxies and work networks block Kimi. Try mobile data to rule out your network.

  5. 5

    Check your limits and credits

    Agent tasks such as Slides, Deep Research and Agent Swarm stop when monthly credits run out.

  6. 6

    Update the app

    On phones and desktop, install the latest Kimi app or Kimi Work version from the official store or kimi.com.

CHECK

Is it you or is Kimi down?

Before you change settings, find out which side the problem is on. It saves a lot of time.

  • 1
    Try another device or network. Open kimi.com on your phone using mobile data. If it works there, the problem is on your side.
  • 2
    Check a status tracker. Third-party monitors such as StatusGator and is-down.ai track Kimi and user reports.
  • 3
    Look for other reports. Search “Kimi down” on X or Reddit. Many reports at once means an outage.
  • 4
    Expect busy spells after launches. Demand jumps when a new model arrives. After K3.1 Pro launched in October 2026, Kimi apologised to subscribers for rate limits.
Does kimi.com load onanother device or network? No Your connectionRouter, VPN, firewall Yes Is there an outage?Check the status page Yes Wait and retryNothing to fix locally No Fix your browserCache, extensions • Restart the router• Turn the VPN off or on• Try mobile data ✓ Fixed Two quick questions tell you whether the problem is onyour side or Kimi’s.

Two questions tell you whether to fix something or simply wait.

FINDER

Find your problem

Pick a category or search, then open a problem to see the fixes.

Showing all 25 problems

Won't loadkimi.com won't load or shows a blank page
  • Refresh the page, then try a private (incognito) window.
  • Clear cookies and cache for kimi.com only.
  • Turn off ad blockers and script-blocking extensions.
  • Try another browser, or mobile data instead of Wi-Fi.
  • If nothing loads anywhere, Kimi may be down. Check a status tracker and try later.
Won't loadKimi doesn't work on my work or school network
  • Your network may block AI sites. Try mobile data to check.
  • Ask your IT team whether kimi.com is allowed.
  • Don't use unofficial mirror sites to get around a block.
Won't loadThe page keeps logging me out
  • Allow cookies for kimi.com in your browser settings.
  • Turn off extensions that delete cookies automatically.
  • Make sure your device's date and time are set automatically.
Busy & slow“Server is busy” or “Please try again later”
  • Wait 30–60 seconds, then send once. Repeated rapid retries can make it worse.
  • Start a new chat, which sends less earlier context.
  • Turn off web search if your question doesn't need it.
  • Try at a quieter time of day.
  • If it continues for hours, check whether others report an outage.
Busy & slowKimi is very slow or stuck “thinking”
  • K3.1 Pro thinks before it answers, so long replies can take a minute or more.
  • Lower the thinking effort for simple tasks.
  • Stop the reply and ask a shorter, more specific question.
  • Start a new chat if the current one is very long.
Busy & slowThe reply stops in the middle
  • Type “continue” and Kimi will usually carry on.
  • Ask for the answer in parts, such as one section at a time.
  • Check your internet connection didn't drop.
Busy & slowAnswers are wrong or made up
  • Ask for sources, page numbers or quotes so you can check.
  • Give Kimi the document or data instead of asking it to recall facts.
  • Add “Tell me if you're not sure” to your prompt.
  • Check important facts yourself. No AI is error-free.
LoginThe login code never arrives
  • Check the country code on your phone number, or the spelling of your email.
  • Wait a minute, then request a new code. Use only the newest code.
  • Look in your spam or junk folder.
  • Try the other sign-in method if your region offers both.
LoginMy phone number isn't accepted
  • Kimi only supports phone numbers in some regions.
  • Virtual (internet) numbers and landlines usually don't work.
  • Sign in with email instead if that option appears.
LoginMy chats or paid plan are missing
  • You may have signed in with a different phone or email, which creates a new empty account.
  • Sign out and sign in again with the method you first used.
  • Paid plans belong to the account that bought them, not the device.
FilesMy file won't upload
  • Each file must be 100 MB or smaller. Compress or split bigger files.
  • You can add up to 50 files at a time.
  • Remove passwords from PDFs and Office files.
  • Rename files with unusual characters in the name.
FilesKimi can't read my PDF or only reads part of it
  • Check the PDF has real text: try selecting some with your mouse.
  • Run OCR on scanned PDFs first.
  • Ask “What is the last heading?” to test if it read the whole file.
  • Split very long files by chapter.
FilesKimi gave text instead of a file (Excel, Word, slides)
  • Use Sheets, Docs or Slides, or the Agent button, not normal chat.
  • Ask for “a downloadable .xlsx (or .docx, .pptx) file”.
  • Check you still have agent credits this month.
Agents & limits“Usage limit reached” or credits used up
  • Agent tasks (Slides, Docs, Sheets, Deep Research, Agent Swarm) use monthly credits.
  • Credits reset each month on your billing date.
  • Normal chat may still work when agent credits run out.
  • Upgrade only from kimi.com or the official app.
Agents & limitsDeep Research or an agent task is stuck
  • Long research can take 10–25 minutes. Leave the tab open.
  • If nothing changes for a long time, stop and start again with a narrower question.
  • Check your credits haven't run out mid-task.
Agents & limitsKimi Slides or Docs looks wrong after export
  • Export to .pptx or .docx, not PDF, if you need to edit.
  • Open the file in a recent version of PowerPoint, Word or WPS.
  • Ask Kimi to fix specific slides with a comment instead of regenerating.
Agents & limitsMy payment or upgrade didn't work
  • Check you're signed in to the same account you paid with.
  • Wait a few minutes, then refresh. Some payments take time to confirm.
  • Contact support with a screenshot of the receipt.
Kimi CodeKimi Code says I've hit my limit
  • Kimi Code has a 5-hour usage window and a weekly limit, depending on your plan.
  • Wait for the window to reset, or switch to a lighter model.
  • Highspeed models use about twice as much quota.
Kimi CodeKimi Code shows a 429 error
  • Read the message: it can mean the server is overloaded or that your quota is used up.
  • For overload, wait and retry. For quota, wait for the reset or upgrade.
  • Avoid running many agents at once.
Kimi CodeMy editor or tool can't connect to Kimi Code
  • Check the base URL: https://api.kimi.ai/coding/v1 (OpenAI-compatible) or https://api.kimi.ai/coding/ (Anthropic-compatible).
  • Use a Kimi Code key, not a Kimi API platform key.
  • Use a current model ID such as k3.1-pro.
API401: Invalid authentication
  • Check the header is Authorization: Bearer YOUR_KEY.
  • Use a key from the same platform as the endpoint. Keys from platform.kimi.ai don't work on other regional platforms.
  • Create a new key if the old one was deleted.
API403 or 429: balance or quota
  • Check your balance in the console and top up.
  • A 403 can also mean the API or model isn't open to your account.
API404: Model not found
  • Check the spelling of the model name, such as kimi-k3.1-pro.
  • K2-series models were closed to new users on 31 October 2026 and retire by the end of December 2026. Move to K3.1 Pro.
API429: Rate limit or engine overloaded
  • Check error.type in the response.
  • engine_overloaded_error: wait and retry with backoff. Topping up won't help.
  • rate_limit_reached_error: reduce concurrency or requests per minute, or move up a tier.
API400: Input exceeds the context length
  • Shorten the input or split the document.
  • Lower max_tokens so prompt plus output fits the model's limit.
  • K3.1 Pro has a 1M-token context window.
MESSAGES

Kimi error messages explained

The exact wording can differ between the website, the apps and over time, but these are the usual meanings.

What you seeWhat it meansWhat to do
Server is busy / try again laterKimi has more requests than it can handle right nowWait a minute, retry once, start a new chat
Usage limit reachedYou've used your plan's allowance for nowWait for the reset or upgrade
Network errorYour connection dropped or something blocked the requestCheck Wi-Fi, VPN and extensions, then refresh
File failed to upload or parseThe file is too big, locked, damaged or a scanUnder 100 MB, no password, OCR for scans
Content can't be shownThe request or reply triggered a safety filterRephrase the request
Context too longThe chat or file is more than the model can holdNew chat, shorter input or split the file
API

Kimi API error codes

From Kimi's API error reference. The JSON response includes error.type, error.message and a request_id to send to support.

CodeMeaningFix
400Bad request: wrong format, missing parameter or input too longCheck the request body and shorten the input
401Invalid or missing API keyCheck the key and use the endpoint for that key's platform
403No permission or balance used upTop up, or check the API is open to your account
404Model or resource not foundCheck the model name and your account tier
429Rate limit, low balance or engine overloadedRead error.type, then back off, reduce concurrency or top up
499Client closed the connectionCheck proxy, KeepAlive and timeout settings
500Server errorRetry later; send the request_id to support if it continues
503Service temporarily unavailableRetry later, usually during scaling or maintenance
Python: retry with backoff
import os, time, random from openai import OpenAI, RateLimitError, APIStatusError client = OpenAI(api_key=os.environ["MOONSHOT_API_KEY"], base_url="https://api.moonshot.ai/v1") def ask(messages, tries=5): for i in range(tries): try: return client.chat.completions.create( model="kimi-k3.1-pro", messages=messages) except RateLimitError as e: # 429 wait = float(e.response.headers.get("retry-after", 2 ** i)) time.sleep(wait + random.random()) except APIStatusError as e: if e.status_code in (500, 503): # server side: retry time.sleep(2 ** i + random.random()) else: # 400/401/403/404: fix the request raise raise RuntimeError("Kimi API still failing after retries")

Retry only what can succeed

  • 1
    Retry: 429 overload and rate limits, 500 and 503. Wait longer each time and respect Retry-After.
  • 2
    Don't retry: 400, 401, 403 and 404. Fix the request, key, balance or model name first.
  • 3
    Log the request_id. Support can find your request with it.

Setup, keys and pricing: Kimi API guide. Coding tools: Kimi Code guide.

SUPPORT

Still not working? Contact Kimi support

On the website

Open Account, then User Feedback, then Help & Feedback in the lower-left corner of kimi.com.

In the app

Go to Settings, then Help & Feedback. Attach a screenshot of the problem.

What to include

Your registered phone or email, the device and browser, the time it happened, a screenshot and, for the API, the request_id.

Stay safe: only use kimi.com and the official apps. Kimi staff will never ask for your login code, and you should only pay inside Kimi. See the Kimi login guide for account help.
FAQ

Kimi troubleshooting FAQ

The most common causes are heavy demand (“server busy”), a browser or network problem, used-up limits or credits, or a temporary outage. Check whether others report problems, then refresh, start a new chat, try another browser and check your limits.

Check a third-party status tracker such as StatusGator or IsDown, or search “Kimi down” on X. If kimi.com fails on several devices and networks and others report problems, it's likely an outage, so wait and try later.

It means Kimi has more requests than it can handle at that moment, often at peak times or right after a new model launch. Wait a minute and retry once, start a new chat, or try at a quieter time.

Kimi K3.1 Pro thinks before answering, so long or hard tasks take longer. Very long chats, big files and busy periods also slow it down. Lower the thinking effort for simple tasks and start new chats often.

Check the country code or email spelling, wait a minute and request a new code, and look in your spam folder. Kimi only supports phone numbers in some regions, and virtual numbers usually don't work.

Kimi Slides, Docs, Sheets and other agent tasks use monthly credits. When they run out, these tasks stop until your credits reset or you upgrade, though normal chat may still work.

A 429 can mean you hit a rate limit, your balance or quota is too low, or the engine is overloaded. Check error.type in the response: back off and retry for overload, reduce concurrency for rate limits, and top up for balance.

On kimi.com, open Account, then User Feedback, then Help & Feedback. In the app, use Help & Feedback in Settings. Include your registered phone or email, a screenshot and the time the problem happened.

Back up and running?

Get more from Kimi with our beginner's guide and ready-to-copy prompts.