AI Reliability Leaderboard
Compare 90-day reliability signals across 800+ AI services: confirmed incidents, community reports, monitoring confidence, and verified availability. Not a model-speed benchmark.
When most AI services have no confirmed hard outage, a single “most reliable” ranking would be misleading. DownForAI shows the underlying reliability signals instead.
817services tracked
134official status sources
90-daywindow
Last updated: Wed, 19 Aug 2026 18:21:02 GMT
Key signals (90 days)
817
services tracked
134
official status sources
100
confirmed incidents (90d)
971
community reports (90d)
717
services with no confirmed outage (90d)
Most reported AI services (90d)
Raw community reports over the last 90 days. Not normalized by user base — popular services may receive more reports.
| Service | Category | Reports 90d | Community-detected outages | Confirmed incidents | Availability | Confidence |
|---|---|---|---|---|---|---|
| Chub AI | Roleplay AI | 131 | 7 | 0 | 100.00% | Basic probe |
| HiWaifu | Roleplay AI | 123 | 12 | 0 | 100.00% | Basic probe |
| Chai AI | Roleplay AI | 112 | 5 | 0 | 100.00% | Basic probe |
| Ollama | LLM | 57 | — | 0 | 100.00% | Basic probe |
| SpicyChat AI | Roleplay AI | 52 | — | 0 | 100.00% | Basic probe |
| Civitai | Image | 47 | 2 | 0 | 100.00% | Status page |
| OpenAI | LLM | 23 | — | 3 | 99.73% | Official API |
| Kiro | Dev Tools | 21 | — | 0 | 100.00% | Basic probe |
| Talkie AI | Roleplay AI | 20 | — | 0 | 100.00% | Basic probe |
| Voicemod | Audio | 19 | — | 0 | 100.00% | Basic probe |
| GitHub Copilot | Dev Tools | 18 | — | 2 | 96.17% | Official API |
| NVIDIA NIM | Infrastructure | 17 | — | 0 | 100.00% | Status page |
| Microsoft Copilot | LLM | 16 | — | 0 | 100.00% | Basic probe |
| Candy AI | Roleplay AI | 15 | 2 | 0 | 100.00% | Basic probe |
| Google Gemini | LLM | 15 | — | 0 | 100.00% | Basic probe |
| JanitorAI | Roleplay AI | 11 | — | 0 | 100.00% | Status page |
| Vast.ai | Infrastructure | 11 | 2 | 0 | 100.00% | Basic probe |
| Moonshot AI (Kimi) | LLM | 10 | — | 7 | 84.03% | Official API |
| Vidu | Video | 9 | — | 0 | 100.00% | Basic probe |
| Anthropic | LLM | 9 | — | 1 | 99.43% | Official API |
Confirmed incidents (90d)
Services with confirmed outage or degradation incidents observed by DownForAI. Incident minutes sum the duration of confirmed incidents.
| Service | Category | Incidents | Incident minutes | Degraded checks | Availability | Source |
|---|---|---|---|---|---|---|
| Elastic AI Search | Search | 21 | 86,312 | 0 | 0.00% | Official API |
| Outranking | Marketing AI | 5 | 21,062 | 0 | 100.00% | Basic probe |
| Moonshot AI (Kimi) | LLM | 7 | 17,537 | 0 | 84.03% | Official API |
| Kajiwoto | Roleplay AI | 3 | 9,605 | 0 | 10.27% | Basic probe |
| Mate AI | Roleplay AI | 3 | 9,258 | 0 | 0.00% | Basic probe |
| ChatGPT | LLM | 3 | 4,470 | 18 | 99.46% | Official API |
| OpenAI Operator | Agents | 3 | 4,470 | 18 | 99.46% | Official API |
| OpenAI Sora | Video | 3 | 4,440 | 21 | 100.00% | Official API |
| Rakuten AI | Productivity | 6 | 4,426 | 0 | 100.00% | Basic probe |
| GPT Image (OpenAI) | Image | 3 | 4,425 | 16 | 100.00% | Official API |
| OpenAI | LLM | 3 | 4,425 | 77 | 99.73% | Official API |
| OpenAI API | Dev Tools | 3 | 4,425 | 18 | 100.00% | Official API |
| Cloudflare Workers AI | Infrastructure | 1 | 3,495 | 181 | 98.91% | Official API |
| ElevenLabs | Audio | 5 | 1,290 | 0 | 100.00% | Official API |
| Retool AI | Agents | 1 | 1,140 | 7 | 100.00% | Official API |
| AI21 Labs | LLM | 1 | 960 | 0 | 100.00% | Official API |
| OpenAI Whisper | Audio | 2 | 855 | 18 | 100.00% | Official API |
| GitHub Models | Dev Tools | 2 | 780 | 3 | 96.17% | Official API |
| GitHub Copilot | Dev Tools | 2 | 736 | 6 | 96.17% | Official API |
| BentoML | Infrastructure | 1 | 614 | 0 | 100.00% | Basic probe |
| Snowflake Cortex | Vector DB | 5 | 599 | 3 | 98.90% | Official API |
| Render AI | Dev Tools | 2 | 359 | 0 | 100.00% | Official API |
| Writecream | Marketing AI | 1 | 300 | 0 | 100.00% | Basic probe |
| Linear | Productivity | 2 | 180 | 1 | 100.00% | Official API |
| Copymatic | Marketing AI | 1 | 180 | 0 | 100.00% | Basic probe |
| Intercom Fin | Support AI | 1 | 180 | 0 | 100.00% | Official API |
| Botpress | Agents | 1 | 180 | 0 | 100.00% | Basic probe |
| DuckDuckGo AI | LLM | 2 | 119 | 0 | 100.00% | Basic probe |
| Airtable AI | Productivity | 1 | 60 | 0 | 100.00% | Official API |
| MarsCode | Dev Tools | 1 | 60 | 0 | 100.00% | Basic probe |
| HeyGen | Video | 1 | 60 | 1 | 100.00% | Official API |
| Canva AI | Image | 1 | 60 | 0 | 100.00% | Official API |
| Claude Chat | LLM | 1 | 60 | 15 | 99.43% | Official API |
| Applitools | Dev Tools | 1 | 60 | 0 | 100.00% | Basic probe |
| Anthropic | LLM | 1 | 60 | 45 | 99.43% | Official API |
Officially monitored reliability leaders
Restricted to the 131 services with an official status API or status page, where availability is measured comparably. Grouped by outcome rather than ranked — most have no confirmed hard outage.
No confirmed incidents or degradations (90d) — 85
Adobe ExpressAlgolia AIAmazon NovaAmazon Q DeveloperAmazon Q DeveloperAssemblyAIAtlassian IntelligenceAWS BedrockAWS SageMakerAzure AI StudioAzure OpenAIBolt.newBrainly AICharacter.AICivitaiClickUp AICodeium / WindsurfComet MLCorcelCrewAICursorDatabricksDatadog LLMDeepgramDeepInfraDescriptDescript VideoDifyFal.aiFal.ai FluxFigma AIFreshdesk AIGammaGitLab DuoGong.ioGoogle AI PlatformGoogle AI StudioGoogle Vertex AIGroqGroq APIHeliconeHubSpot AIHugging FaceIdeogramJanitorAIJasper AILambda LabsLangfuseLe Chat (Mistral)LovableLuma AILuma Dream MachineMidjourneyMintlifyMiro AIMistral AIModalMonday AINeon pgvectorNVIDIA NIMOpenRouterPaperspacePerplexityPineconePoePortkeyRailway AIReplicateResemble AIRev AIRunwaySalesforce EinsteinSlack AISourcegraph CodyStable AudioStable Video DiffusionSynthesiaTabnineTalkdesk AITogether AIUdioVal TownVoiceflowWritesonicZendesk AI
Confirmed degradations / incidents, no hard outage — 23
| Service | Source | Availability | Incidents | Degraded checks | Reports |
|---|---|---|---|---|---|
| Couchbase Capella | Official API | 100.00% | 0 | 181 | 0 |
| Zapier AI | Official API | 100.00% | 0 | 25 | 0 |
| OpenAI Sora | Official API | 100.00% | 3 | 21 | 0 |
| OpenAI Whisper | Official API | 100.00% | 2 | 18 | 0 |
| OpenAI API | Official API | 100.00% | 3 | 18 | 0 |
| GPT Image (OpenAI) | Official API | 100.00% | 3 | 16 | 0 |
| Retool AI | Official API | 100.00% | 1 | 7 | 0 |
| v0 by Vercel | Official API | 100.00% | 0 | 7 | 0 |
| Vercel | Official API | 100.00% | 0 | 7 | 0 |
| Zoom AI Companion | Official API | 100.00% | 0 | 6 | 0 |
| Snyk Code AI | Official API | 100.00% | 0 | 5 | 0 |
| Grammarly AI | Official API | 100.00% | 0 | 4 | 0 |
| MongoDB Atlas Vector | Official API | 100.00% | 0 | 4 | 0 |
| HeyGen | Official API | 100.00% | 1 | 1 | 0 |
| StudyFetch | Official API | 100.00% | 0 | 1 | 0 |
| Linear | Official API | 100.00% | 2 | 1 | 0 |
| Make (ex-Integromat) | Official API | 100.00% | 0 | 1 | 0 |
| Airtable AI | Official API | 100.00% | 1 | 0 | 0 |
| AI21 Labs | Official API | 100.00% | 1 | 0 | 0 |
| Render AI | Official API | 100.00% | 2 | 0 | 0 |
| Canva AI | Official API | 100.00% | 1 | 0 | 0 |
| ElevenLabs | Official API | 100.00% | 5 | 0 | 0 |
| Intercom Fin | Official API | 100.00% | 1 | 0 | 0 |
Confirmed outages (90d) — 23
| Service | Source | Availability | Incidents | Degraded checks | Reports |
|---|---|---|---|---|---|
| Elastic AI Search | Official API | 0.00% | 21 | 0 | 0 |
| Moonshot AI (Kimi) | Official API | 84.03% | 7 | 0 | 10 |
| GitHub Copilot | Official API | 96.17% | 2 | 6 | 18 |
| GitHub Models | Official API | 96.17% | 2 | 3 | 0 |
| Anthropic | Official API | 99.43% | 1 | 45 | 9 |
| Cohere | Official API | 99.45% | 0 | 0 | 0 |
| Snowflake Cortex | Official API | 98.90% | 5 | 3 | 1 |
| Cloudflare Workers AI | Official API | 98.91% | 1 | 181 | 0 |
| OpenAI | Official API | 99.73% | 3 | 77 | 23 |
| n8n | Status page | 99.46% | 0 | 0 | 0 |
| Notion AI | Official API | 99.46% | 0 | 0 | 1 |
| ChatGPT | Official API | 99.46% | 3 | 18 | 5 |
| Fireworks AI | Status page | 99.45% | 0 | 0 | 0 |
| OpenAI Operator | Official API | 99.46% | 3 | 18 | 0 |
| Humanloop | Status page | 99.46% | 0 | 0 | 0 |
| Anthropic API | Official API | 99.43% | 0 | 13 | 1 |
| Stability AI | Official API | 99.69% | 0 | 0 | 0 |
| Supabase | Official API | 99.73% | 0 | 271 | 0 |
| Cartesia | Status page | 99.46% | 0 | 0 | 0 |
| Inflection Pi | Official API | 99.45% | 0 | 0 | 0 |
| Claude Chat | Official API | 99.43% | 1 | 15 | 4 |
| Otter.ai | Status page | 99.46% | 0 | 0 | 1 |
| Supabase Vector | Official API | 99.45% | 0 | 135 | 0 |
Reliability leaders by category
Browse like-for-like reliability rankings within each category.
LLM · 58 tracked
View full LLM reliability rankings →Image · 54 tracked
View full Image reliability rankings →Video · 52 tracked
View full Video reliability rankings →Audio · 49 tracked
View full Audio reliability rankings →Dev Tools · 74 tracked
View full Dev Tools reliability rankings →Infrastructure · 48 tracked
View full Infrastructure reliability rankings →Search · 23 tracked
View full Search reliability rankings →Productivity · 55 tracked
View full Productivity reliability rankings →Browse all categories →
How reliable is our signal?
Monitoring confidence reflects how a service is observed — the quality of our measurement, not the service's reliability.
55Official API
79Status page
669Basic probe
12Limited
2Unverifiable
How this leaderboard works
- Availability = non-outage rate over a rolling 90-day window. “Down” means a hard OUTAGE only. A degraded period is not downtime — it is counted as a separate signal. A blocked or rate-limited probe is never counted as an outage.
- Confirmed incidents exclude false positives. Community reports are user-submitted, counted as raw totals not normalized by user base, and shown as their own independent signal — a popular service naturally receives more.
- Monitoring confidence (official API, status page, basic probe, limited, unverifiable) describes the quality of our measurement, not the service's reliability — we never rank by it.
- Each surface is re-checked roughly every 75 minutes. Any response-time figures shown elsewhere reflect the monitored surface (often a homepage or status page) — not model inference speed or tokens-per-second.