Replicate SLA
Uptime guarantee vs measured reliability
Official SLA Commitment
Applies to
Enterprise plan only
Scope
Enterprise contracts with negotiated performance SLAs
Measurement
Uptime and performance commitments defined in offline enterprise agreement; no public percentage published
Exclusions
Self-serve pay-as-you-go usage; public and private model inference on shared infrastructure
Not covered
Standard pay-as-you-go API access, public model runs, and private models on shared queues
24h Uptime
100.00%
7d Uptime
100.00%
30d Uptime
100.00%
Pricing & SLA by Tier
| Plan | SLA Guarantee | Price | Meeting 30d? |
|---|---|---|---|
| Pay-as-you-go (public models) | No uptime SLA | Per-second GPU/CPU billing (e.g. T4 GPU $0.000225/sec; token-priced models vary) | — |
| Private models (dedicated hardware) | No uptime SLA | Billed for setup, idle, and active time on dedicated instances | — |
| Enterprise | No uptime SLA | Custom (volume discounts, committed spend, performance SLAs via sales) | — |
Source: replicate.com/pricing
Industry SLA Tier Compliance
| Tier | Downtime Budget/Month | 24h | 7d | 30d |
|---|---|---|---|---|
| 99.9%(three nines) | 43m 12s | |||
| 99.95% | 21m 36s | |||
| 99.99%(four nines) | 4m 19s | |||
| 99.999%(five nines) | 26s |
Equivalent Downtime
Last 24 hours
< 1m
Last 7 days
< 1m
Last 30 days
< 1m
Incident Summary (90 days)
Total Incidents
6
Active Now
0
Resolved
6
Avg. Resolution
5h 45m
Replicate SLA: What You Need to Know
Replicate guarantees uptime measured as enterprise contracts with negotiated performance slas. This commitment applies exclusively to Enterprise plan customers — other tiers have no contractual uptime guarantee.
Based on DevHelm's continuous monitoring, Replicate currently shows a 30-day measured uptime of 100.00%. Uptime data is still being collected for a complete assessment.
What's Covered (and What Isn't)
Replicate's SLA specifically covers enterprise contracts with negotiated performance slas. It explicitly excludes: Standard pay-as-you-go API access, public model runs, and private models on shared queues. This means outages in excluded components don't count toward SLA credits.
Impact on Your Application
At 100.00% uptime, you can expect approximately < 1m of cumulative downtime per year from Replicate alone. If your application depends on additional services, their independent failure probabilities compound — making dependency monitoring essential for maintaining your own SLA commitments.
Monitoring Methodology
DevHelm tracks Replicate's official status page continuously and aggregates component-level status data into uptime percentages. Data is refreshed every 5 minutes. Uptime is calculated as the percentage of time all components report operational status.
