AI Infrastructure Overview
Live reliability, latency, and incident tracking across major AI providers. Public data is updated frequently and viewable without an account.
Azure OpenAIOperational
P95 latency
910 ms
Error 0.60%
GroqPartial Outage
P95 latency
740 ms
Error 2.40%
AnthropicOperational
P95 latency
970 ms
Error 0.50%
GoogleOperational
P95 latency
1010 ms
Error 0.70%
MistralOperational
P95 latency
960 ms
Error 0.80%
Provider leaderboard
Top providers ranked by reliability score.
| Provider | Reliability | P95 latency | Error rate | Rate limit | Incidents (7d) | Trend |
|---|---|---|---|---|---|---|
Azure OpenAIOperational | 92/100 | 910 ms | 0.60% | 0.80% | 0 | |
GroqPartial Outage | 92/100 | 740 ms | 2.40% | 6.20% | 0 | |
AnthropicOperational | 91/100 | 970 ms | 0.50% | 0.70% | 0 | |
GoogleOperational | 91/100 | 1010 ms | 0.70% | 0.60% | 0 | |
MistralOperational | 91/100 | 960 ms | 0.80% | 1.00% | 0 | |
CohereOperational | 90/100 | 1030 ms | 0.70% | 0.90% | 0 | |
OpenAIDegraded | 87/100 | 1320 ms | 1.60% | 2.10% | 0 | |
AWS BedrockDegraded | 85/100 | 1490 ms | 1.60% | 1.50% | 0 |
Model radar
Top models by reliability score. Updated 2025-12-30 07:20:21 UTC.
| Model | Provider | Reliability | P95 latency | Error rate | Rate limit | Trend |
|---|---|---|---|---|---|---|
| Llama 3.1 8B (Groq) | Groq | 98/100 | 320 ms | 0.35% | 0.90% | |
| Llama 3.3 70B (Groq) | Groq | 97/100 | 450 ms | 0.40% | 1.20% | |
| Claude 3 Haiku | Anthropic | 96/100 | 600 ms | 0.08% | 0.40% | |
| Gemini 2.0 Flash-Lite | 95/100 | 650 ms | 0.22% | 0.60% | ||
| GPT-4o mini | OpenAI | 94/100 | 700 ms | 0.10% | 0.60% | |
| Claude 3.5 Sonnet | Anthropic | 93/100 | 820 ms | 0.10% | 0.50% |
Operational
5
Degraded
2
Active incidents
2
Median P95
1010 ms
Active incidents
View allProvider: OpenAI | Started: 2025-12-27 12:04:00 UTC
Provider: Groq | Started: 2025-12-12 16:40:00 UTC
Why teams use CapPulse
Public dashboards show the truth. Paid plans unlock alerting, routing policies, and long-term retention.
Sponsor
Monitoring partner slot. Sponsorships are labeled and never influence rankings.
Request media kit ->Get incident updates in your inbox.
Receive a daily or weekly digest of AI provider incidents. Slack and webhooks are available on paid plans.