vincentcouey/ai-api-reliability-2026

AI API Reliability 2026: what AI providers declare on their own status pages

First-party research dataset from the Lattice network, published open under CC-BY 4.0 with a permanent DOI. Nothing here is scraped from another dataset: it is computed and published under a single ORCID-verified byline.

DOI	10.5281/zenodo.23724799
Published by	Nesyona (nesyona.com)
Study page	https://nesyona.com/research/ai-api-reliability-2026/
Licence	CC-BY 4.0, reuse freely with attribution
Publication date	2026-10-28

What it is

A cross-provider ledger of 1,510 incidents that 11 AI API providers (OpenAI, Anthropic, Google Cloud, Mistral, Groq, Cohere, Perplexity, Fireworks AI, Cerebras, Stability AI, ElevenLabs) declare on their own public status pages, captured on 2026-10-28. Every figure is what the provider DECLARED, not an independent uptime measurement. Components are mapped to one taxonomy by published keyword rules; each provider carries its own coverage window (8 providers expose full history, 3 expose only their newest incidents and gain depth from weekly captures; a window not covered is null, never zero). Includes SLA terms exactly as read from provider pages on the capture date, a revision ledger (empty in version 1 by construction), a one-time zero-token probe burst from a single machine (a snapshot, not uptime) and the URL, read time and SHA-256 of every archived raw status-page response. Re-captur

Citation

@dataset{ai_api_reliability_2026,
  title  = {AI API Reliability 2026: what AI providers declare on their own status pages},
  author = {Couey, Vincent Wesley},
  year   = {2026},
  doi    = {10.5281/zenodo.23724799},
  url    = {https://doi.org/10.5281/zenodo.23724799}
}

Provenance

Part of the Lattice open research index. Every study is CC-BY, DOI-backed and published under one byline. The full index: https://www.deepsynthesis.org/#research