Skip to main content
deepseek.health
Independent monitoring tool. deepseek.health is not affiliated with, operated by, or endorsed by DeepSeek, Mistral, Hugging Face, or any provider listed. Official DeepSeek status: status.deepseek.com.
DeepSeek

Is DeepSeek down? Live API status

Open-weight chat & reasoning models, OpenAI-compatible

Live DeepSeek status — API, console, docs and marketing endpoints checked from your browser against api.deepseek.com, platform.deepseek.com, api-docs.deepseek.com and www.deepseek.com.

DeepSeek status can be verified live by testing direct connectivity from your browser to DeepSeek's public endpoints: api.deepseek.com, platform.deepseek.com, api-docs.deepseek.com and www.deepseek.com. Reachability and latency of DeepSeek public web endpoints (API, console, docs and marketing site) are checked live from your browser. An HTTP 429 from the API usually means your account has reached its concurrent-request limit for that model, while a 503 signals server overload; neither by itself means a complete outage.

DeepSeek API Status

Live, in-browser checks against DeepSeek's public APIs. Refreshes every 15s.

Checking…

API

Reachability of api.deepseek.com (auth-gated — DNS/TCP/TLS + 401 shape)

3 endpoints
  • API root
    api.deepseek.com — chat & reasoning inference edge
    —ms
    warming up
  • Chat completions
    OpenAI-compatible /v1/chat/completions endpoint
    —ms
    warming up
  • Models list
    /v1/models catalog endpoint
    —ms
    warming up

Platform

Console, docs and marketing front-ends

3 endpoints
  • Platform console
    platform.deepseek.com — API keys & billing
    —ms
    warming up
  • API docs
    api-docs.deepseek.com
    —ms
    warming up
  • Marketing site
    www.deepseek.com
    —ms
    warming up
Polling every 15s · Last updated — · Next in 15sOpen DeepSeek ↗

DeepSeek's API requires a bearer token and does not return permissive CORS headers, so browser checks confirm DNS, TCP and TLS reachability only. For live incident + capacity flags, see DeepSeek's official status page embedded below.

Sponsored · DeepSeek
Ask Deep
LLM · $0.14/M

Hey, I'm Deep — your DeepSeek guide. Ask me about pricing, models, rate limits, or how to swap in the API using the OpenAI SDK.

Open DeepSeek platform

Sponsored content. deepseek.health may earn a referral reward when you sign up via this widget. Answers are informational only — not investment or engineering advice.

Official DeepSeek status page

Open ↗

About DeepSeek

DeepSeek is a Chinese AI lab whose open-weight and hosted models (DeepSeek-V3, DeepSeek-R1) offer OpenAI-compatible chat and reasoning APIs at a fraction of the cost of GPT-4o. This page probes DeepSeek's public API, console, docs and marketing endpoints on every poll to confirm they're reachable from your network.

DeepSeek at a glance

Founded
2023
HQ
Hangzhou, China
Flagship models
DeepSeek-V3, DeepSeek-R1
API compat
OpenAI /v1/chat/completions
Pricing
From $0.14 / 1M input tokens (cached)
Context window
64K (chat), 64K (reasoner)

Common DeepSeek issues

  • Empty tokens / hanging streams

    Under heavy load DeepSeek keeps the connection open while a request waits: non-streaming responses receive blank lines and streaming responses receive SSE ': keep-alive' comments. If inference has not started after 10 minutes, the server closes the connection. Parse around these keep-alives, set a finite client deadline, and never automatically replay a stream that already delivered tokens.

  • 429 Too Many Requests

    The limit is concurrent in-flight requests per account and model. Multiple API keys share the account's concurrency, so splitting keys does not help. Pace or queue requests, use bounded backoff with jitter, and request a capacity increase if needed.

    Read the DeepSeek rate limits & retries guide →
  • Reasoner returns empty content

    deepseek-reasoner streams its chain-of-thought in reasoning_content before final content. If you only read content, early tokens look empty — that's expected, not an outage.

Background

DeepSeek shipped V2 in mid-2024 at a price point that undercut GPT-4-class models by 10-50x. V3 (December 2024) matched GPT-4o on many benchmarks, and R1 (January 2025) demonstrated that a hosted open-weight reasoning model could rival OpenAI o1 for a small fraction of the cost — briefly moving public markets and reshaping API pricing across the industry.

Frequently asked

What is the difference between chat.deepseek.com and the DeepSeek API?
chat.deepseek.com is the consumer web interface, while api.deepseek.com powers developer integrations. When the web chat shows 'Server is busy', the API often remains accessible via OpenAI-compatible endpoints or third-party providers.
How do I check if DeepSeek R1 or V3 is operational?
Both DeepSeek-V3 (deepseek-chat) and DeepSeek-R1 (deepseek-reasoner) share the api.deepseek.com gateway. You can verify operational status by pinging the /v1/models or /v1/chat/completions endpoints using an active API key.
How can I check if chat.deepseek.com is working?
deepseek.health does not check chat.deepseek.com — the dashboard probes DeepSeek's public endpoints: api.deepseek.com, platform.deepseek.com, api-docs.deepseek.com and www.deepseek.com. Open the chat site in your own browser, or check DeepSeek's official status page at status.deepseek.com.
What should I do if DeepSeek returns 429 or 503 errors?
A 429 means your account has too many requests in flight for that model: pace or queue new requests and retry with bounded, jittered backoff. A 503 means DeepSeek's servers are overloaded: wait briefly, then retry a limited number of times. If you see 402, that is an insufficient balance, not a rate limit — top up rather than retry. See the deepseek.health rate limits & retries guide for a full decision table.
Is the DeepSeek API down right now?
You can check current DeepSeek API reachability on deepseek.health: the dashboard tests api.deepseek.com, platform.deepseek.com, api-docs.deepseek.com and www.deepseek.com from your browser, including the chat-completions and models endpoints. For confirmed incidents, see DeepSeek's official status page at status.deepseek.com.
Why is DeepSeek returning 500 or 503 errors?
According to DeepSeek's error-code docs, 500 is a server-side issue and 503 means the server is overloaded. Wait briefly and retry a bounded number of times with jittered backoff. A fallback provider can reduce impact, but no retry or fallback strategy guarantees zero interruption.
How can I test DeepSeek API availability in code?
Send an authenticated GET or POST request to https://api.deepseek.com/models or https://api.deepseek.com/v1/chat/completions. A 200 OK HTTP response indicates the API is operational; 429 indicates rate limiting, and 5xx indicates upstream service degradation.
Is DeepSeek down right now?
The dashboard above polls DeepSeek's public endpoints — api.deepseek.com, platform.deepseek.com, api-docs.deepseek.com and www.deepseek.com — from your browser every 15 seconds. Green dots mean those endpoints are reachable from your network, not that every request will succeed. For internal capacity or model-specific incidents, check the embedded status.deepseek.com panel.
Why do I keep getting 429s from DeepSeek?
DeepSeek limits concurrent in-flight requests per account and model (counted from send until the response completes), not a fixed requests-per-minute number. All API keys on an account share that limit. Track pending requests, queue new ones when you near the ceiling, retry 429s with bounded backoff, and ask DeepSeek for a capacity increase if you consistently need more.
Is the API OpenAI-compatible?
Yes. Swap base_url to https://api.deepseek.com and you can reuse the OpenAI SDK for chat and streaming. Function calling and JSON mode are supported on the chat model.
What models are live?
deepseek-chat (the V3 chat model) and deepseek-reasoner (the R1-series reasoning model). Older snapshots are usually deprecated on 30-day notice.

About this check

All data is provided on a best-effort basis with no guarantees about accuracy or availability. This is an unofficial status page and is not affiliated with, endorsed by, or sponsored by DeepSeek.

Checks run client-side from your browser against DeepSeek's public endpoints. Latency reflects your network path, not the server's health alone. See our resources for more on how to read these checks.

All tracked services

Live checks across LLM APIs and the ML hubs that host them.

LLM APIs

DeepSeek

Open-weight chat & reasoning models, OpenAI-compatible

View status →
Mistral

European LLM lab · open + hosted models

View status →
OpenAI-compat

Drop-in providers speaking /v1/chat/completions

View status →

ML hubs

Hugging Face

Open-source ML hub · models · datasets · Spaces

View status →