Amazon Bedrock SLAAWS · AI models · AWS · AI models
Promise vs reality
Downtime used against the allowance
3 d of major incidents, 8.2x the allowance8.8 h allowed in 365 days
3 major and 0 minor incidents; worst region me-central-1. 0 min degraded (not counted).
Major incident minutes per month
Allowed per month: 44 min
- Promised
- 99.9%
- Observed, 365d
- 99.178%
- Major incident time
- 3 d
- Credit on first breach
- 10%
Commitments and credits
- StandardAmazon Bedrock APIs for models available on Bedrock, per AWS region99.9%
Monthly uptime below, credit of the bill
- Below 99.9%10%
- Below 99%25%
- Below 95%100%
Terms
- Measured
- Average of 5-minute Availability (share of Bedrock API requests not returning a 500 error) per AWS region
- Credit cap
- Up to 100% of Bedrock charges in the affected region at the lowest tier; issued only if over $1
- Claim window
- By the end of the second billing cycle after which the incident occurred
- How to claim
- Open a case in the AWS Support Center with 'SLA Credit Request' in the subject line and request logs
Not covered
- Factors outside AWS's reasonable control
- Customer actions or inactions (e.g. provisioned throughput scaling, VPC/credential misconfiguration)
- Not following Bedrock documentation best practices
- Customer equipment, software or other technology
- Underlying software causing repeated model crashes or an inoperable model
Monthly Uptime Percentage ... Less than 99.9% but greater than or equal to 99.0% 10% ... Less than 99.0% but greater than or equal to 95.0% 25% ... Less than 95.0% 100%
The commitment is expressed only via the credit table ('Monthly Uptime Percentages set forth in the table below'); 99.9% is the table's top threshold. No latency/throughput SLA is published.
Evidence: incidents, last 365 days
Other ai models SLAs
ServiceDowntime vs allowed
Amazon BedrockAWS99.9% / 99.178%
Vertex AI (Gemini API on Vertex)Google Cloud99.9% / 100%
Azure OpenAI in Foundry ModelsMicrosoft Azure99.9% / n/a
Claude API Priority TierAnthropic99.5% / 98.8%
Cohere API (SaaS: Command, Embed, Rerank)Coherenone / 100%
Hugging Face Inference EndpointsHugging Facenone / n/aNo public SLA
Promised in the SLAObserved on the vendor's status pagePartial historyShortfall
Summaries of published SLAs; the contract you sign governs. Logos via logo.dev; trademarks belong to their owners.