WhatsDownNow
For AI builders

Is your AI provider down? Here's who else is up.

When ChatGPT, Claude, Gemini or another model API stumbles, you need a working alternative — not a forum thread. This board watches the AI providers already in our catalog, shows whether each one is up, and suggests healthy fallbacks your app or agent can switch to.

1 of 14 AI APIs we watch is having problems right now. Healthy alternatives are listed next to each one.

AI API providers

These are model and inference APIs you might actually swap between. Verdicts combine each provider's official status page (when they publish one) with our own HTTP check, about every 11 minutes.

ChatGPTUp

Official status: all systems operational; our check: OK 200 in 213ms

7-day uptime
98.3%
Probe p50 / p95
217 ms / 350 ms
Latest probe
213 ms
Incidents (7d)
1
Healthy — no fallback needed
ClaudeUp

Official status: all systems operational; our check: OK 401 in 113ms (reachable; API answered and asked for a key, as expected)

7-day uptime
53.5%
Probe p50 / p95
108 ms / 250 ms
Latest probe
113 ms
Incidents (7d)
2
Healthy — no fallback needed
GeminiUp

Official status: no active incidents; our check: OK 200 in 172ms

7-day uptime
100.00%
Probe p50 / p95
163 ms / 900 ms
Latest probe
172 ms
Incidents (7d)
0
Healthy — no fallback needed
Mistral AIDegraded

Official status: Vibe Work issues with GLM 5.3 Short; our check: OK 403 in 472ms (reachable; site blocks automated checks)

7-day uptime
85.4%
Probe p50 / p95
125 ms / 500 ms
Latest probe
472 ms
Incidents (7d)
1
Try insteadOpenRouterPerplexityHugging Face
GroqUp

Official status: all systems operational; our check: OK 200 in 92ms

7-day uptime
100.00%
Probe p50 / p95
125 ms / 500 ms
Latest probe
92 ms
Incidents (7d)
0
Healthy — no fallback needed
PerplexityUp

Official status: all systems operational; our check: OK 403 in 5ms (reachable; site blocks automated checks)

7-day uptime
100.00%
Probe p50 / p95
22 ms / 47 ms
Latest probe
5 ms
Incidents (7d)
0
Healthy — no fallback needed
CohereUp

Official status: all systems operational; our check: OK 200 in 322ms

7-day uptime
100.00%
Probe p50 / p95
313 ms / 500 ms
Latest probe
322 ms
Incidents (7d)
0
Healthy — no fallback needed
GrokUp

Official status: no public status feed; our check: OK 200 in 359ms

7-day uptime
100.00%
Probe p50 / p95
388 ms / 700 ms
Latest probe
359 ms
Incidents (7d)
0
Healthy — no fallback needed
DeepSeekUp

Official status: no public status feed; our check: OK 200 in 420ms

7-day uptime
100.00%
Probe p50 / p95
425 ms / 575 ms
Latest probe
420 ms
Incidents (7d)
0
Healthy — no fallback needed
Together AIUp

Official status: no public status feed; our check: OK 200 in 85ms

7-day uptime
100.00%
Probe p50 / p95
35 ms / 88 ms
Latest probe
85 ms
Incidents (7d)
0
Healthy — no fallback needed
ReplicateUp

Official status: no public status feed; our check: OK 200 in 709ms

7-day uptime
100.00%
Probe p50 / p95
550 ms / 1250 ms
Latest probe
709 ms
Incidents (7d)
0
Healthy — no fallback needed
Hugging FaceUp

Official status: no recent incidents in official feed; our check: OK 200 in 33ms

7-day uptime
100.00%
Probe p50 / p95
44 ms / 125 ms
Latest probe
33 ms
Incidents (7d)
0
Healthy — no fallback needed
OpenRouterUp

Official status: no public status feed; our check: OK 200 in 70ms

7-day uptime
100.00%
Probe p50 / p95
65 ms / 95 ms
Latest probe
70 ms
Incidents (7d)
0
Healthy — no fallback needed
Fireworks AIUp

Official status: all systems operational; our check: OK 200 in 225ms

7-day uptime
100.00%
Probe p50 / p95
258 ms / 500 ms
Latest probe
225 ms
Incidents (7d)
0
Healthy — no fallback needed

Latency is from our HTTP probe of the provider's public website, API endpoint, or status page — not model inference latency. Uptime is the share of time our verdict was up, for the window we have been recording (monitoring began mid-November 2026). Probe history fills in as checks run.

Fallback API

Agents and apps can ask “what's a healthy alternative to this provider?” without scraping this page.

curl https://whatsdownnow.com/api/v1/ai/health
curl "https://whatsdownnow.com/api/v1/ai/fallback?provider=chatgpt"
curl "https://whatsdownnow.com/api/v1/services/claude/history?days=7"

From an MCP client, call get_ai_fallback (omit provider for the whole board). get_status also includes fallback hints when the service is an AI API.

Other AI tools we watch

IDEs, image and voice apps, and builders. They are not treated as interchangeable fallbacks for the APIs above.

How to read this

  • We do not call paid model APIs. A green probe means the public endpoint answered our request, not that a particular model or region is healthy.
  • Official status pages can lag, and so can we. Confirm a critical decision on the provider's own status page.
  • Want alerts in Slack, Discord, email or a signed webhook when a provider goes down? Create a stack or let your agent do it.
  • Service names are trademarks of their owners. WhatsDownNow is not affiliated with any of them.