Uptimeby Glyph

Promised uptime vs actual

Vendors35 Services99

35 vendors · 99 services

OpenAI API Scale Tier SLA

OpenAI · AI models · promises up to 99.9% · checked 47d ago

Official SLA Status page

No major outage in the 58 days of history: 100% observed against the 99.9% Scale Tier: GPT-5.5.

Last 90 days: no major outage.

Incidents, last 365 days

MinorOct 27, 2026

6.1 Sol Degraded Model Performance

A minor incident on Batch, Chat Completions, Responses for 78 min, not counted as downtime.

MinorOct 22, 2026

Elevated ChatGPT conversation and Responses API errors

A minor incident on Conversations, Responses for 6.5 h, not counted as downtime.

MinorOct 21, 2026

Elevated error rates for gpt-5.6-luna API requests

A minor incident on Chat Completions, Responses for 69 min, not counted as downtime.

MinorOct 21, 2026

Degraded performance in ChatGPT, Codex, and APIs

A minor incident on Connectors/Apps, CLI, Codex in ChatGPT Desktop and 3 more for 22 min, not counted as downtime.

MinorOct 19, 2026

File Download errors

A minor incident on File uploads, Images, Files and 2 more for 42 min, not counted as downtime.

MinorOct 6, 2026

Errors processing large PDFs in the Responses API

A minor incident on Responses for 13 h, not counted as downtime.

MinorSep 30, 2026

Elevated latency for some API requests

A minor incident on Responses, Chat Completions for 3.5 h, not counted as downtime.

MinorSep 29, 2026

Elevated errors across ChatGPT, Codex, and the API including the Agents API

A minor incident on Login, Compliance API, File uploads and 3 more for 5.3 h, not counted as downtime.

MinorSep 17, 2026

We are seeing elevated error rates across API models

A minor incident on Images, Batch, Chat Completions and 3 more for 60 min, not counted as downtime.

MinorSep 11, 2026

Elevated errors for GPT-5.6 Sol on the API

A minor incident on Chat Completions, Responses for 2.2 h, not counted as downtime.

Major outage time per month

Allowed per month: 44 min

Commitments and credits

99.9%

Scale Tier: GPT-5.5

Scale Tier token units for GPT-5.5 (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 50 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-5.4 mini

Scale Tier token units for GPT-5.4 mini (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 100 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-5.4 (excludes long-context >272K)

Scale Tier token units for GPT-5.4 (excludes long-context >272K) (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 50 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-5.2

Scale Tier token units for GPT-5.2 (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 50 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-5.1

Scale Tier token units for GPT-5.1 (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 50 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-5

Scale Tier token units for GPT-5 (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 50 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-5 mini

Scale Tier token units for GPT-5 mini (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 80 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-4.1 (excludes long-context >128K)

Scale Tier token units for GPT-4.1 (excludes long-context >128K) (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 80 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-4.1 mini (excludes long-context >128K)

Scale Tier token units for GPT-4.1 mini (excludes long-context >128K) (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 90 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-4.1 nano (excludes long-context >128K)

Scale Tier token units for GPT-4.1 nano (excludes long-context >128K) (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 100 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-4.1 fine tuning

Scale Tier token units for GPT-4.1 fine tuning (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 80 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-4.1 mini fine tuning

Scale Tier token units for GPT-4.1 mini fine tuning (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 90 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: o3

Scale Tier token units for o3 (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 80 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: o4-mini

Scale Tier token units for o4-mini (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 90 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-4o

Scale Tier token units for GPT-4o (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 80 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-4o mini

Scale Tier token units for GPT-4o mini (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 90 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: GPT-4o mini fine tuning

Scale Tier token units for GPT-4o mini fine tuning (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 90 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: o1

Scale Tier token units for o1 (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 80 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

99.9%

Scale Tier: o3-mini

Scale Tier token units for o3-mini (Enterprise customers, min 30-day purchase). Latency SLA: 99% > 90 tokens per second, calculated as p50 request latency on a per 5 minute basis.

Credit: No credit schedule published

SLA changes

From the change logSep 23, 2026

OpenAI API Scale Tier

GPT-5.5 Scale Tier latency SLA listed as '99% > 100 tokens per second' in the 2026-06-11 snapshot; the live page on 2026-09-23 lists '99% > 50 tokens per second'. Exact change date unknown (between those dates).

From the change logAug 10, 2025

OpenAI API Scale Tier

Latency SLA methodology changed from 'average request latency on per minute basis across the month' (e.g. GPT-4o 99% > 25 tokens/s in Aug 2024) to 'p50 request latency on a per 5 minute basis'. Change occurred between the 2025-04-04 and 2025-08-10 archived snapshots; uptime stayed 99.9%.

More in ai models

Missed its SLA99.9%

AWS promises 99.9% uptime, with a 10% credit when it misses; its status page showed 3 d of major outage in the last year.

Observed 99.178% over 365 days

Claude API (Standard tier)

Anthropic publishes no SLA for it; its status page showed 4.4 d of major outage in the last year.

Observed 98.805% over 365 days

Missed its SLA99.5%

Claude API Priority Tier

Anthropic promises 99.5% uptime; its status page showed 4.3 d of major outage in the last year.

Observed 98.809% over 365 days

Claude Enterprise (claude.ai)

Anthropic publishes no SLA for it; its status page showed 4.5 d of major outage in the last year.

Observed 98.757% over 365 days

Cohere API (SaaS: Command, Embed, Rerank)

Cohere publishes no SLA for it; its status page showed no major outage in the last year.

Observed 100% over 365 days

Cohere private deployments / Model Vault / North

Cohere publishes no SLA for it; its status history is still building.

Gemini API (Google AI Studio)

Google publishes no SLA for it; its status page has no history this site can read.

Kept its SLA99.5%

Gemini API on Vertex AI

Google Cloud promises 99.5% uptime, with a 10% credit when it misses; its status page showed no major outage in the 252 days on file.

Observed 100% over 365 days

ServicePromisedFirst creditObserved 365d

Terms, exclusions and sources

Measured: Uptime window not stated; credits are for 'the calendar month of that Scale Tier token unit purchase'. Latency SLA calculated as p50 request latency on a per 5 minute basis (older enterprise agreements: per minute).
Not covered: Credit amounts / schedule not published; Long-context requests excluded for GPT-4.1 family (>128K) and GPT-5.4 (>272K); Only models released before GPT-5.6 (newer models use Reserved Tier); Enterprise customers only; via sales order form
Status history: Status history since Sep 9, 2026
Weighted uptime: All incidents, weighted: 99.122% · Downtime

Higher reliability: Scale Tier traffic offers a 99.9% uptime SLA and prioritized compute. ... GPT-5.5 50,000 TPM $750.00 per unit/day N/A 99.9% 99% > 50 tokens per second

OpenAI, SLA. Read the SLA

Published: 99.9% uptime SLA plus per-model latency SLA for every listed model. FAQ: 'What happens if the latency and uptime SLA are both violated? You will be credited with the greater of the two SLA amounts for the calendar month' - but the credit amounts themselves are not published. Page says 'Scale Tier is available on models released before GPT-5.6. For GPT-5.6 and future model releases, see Reserved Tier'. Reserved Tier FAQ (https://openai.com/api-reserved-tier/): 'SLA for the service tier you use for your requests, such as Fast mode or Standard, will apply' - no separate number. openai.com pages were read via a real browser (curl is blocked by Cloudflare).

All OpenAI SLAs How this is measured