Verified history
Perplexity raised Agent API per-minute limits for Tiers 4 and 5
The rate-limits page now lists 4,000 RPM at Tier 4 and 8,000 RPM at Tier 5 (previously 2,000 RPM for both) alongside unchanged QPS values. No effective date is published.
High-tier accounts can sustain up to four times the previous per-minute Agent throughput.
Perplexity now meters the Search API in query units
The Search API limit is expressed as 50 query units per second with a 50-unit burst; a request containing five queries consumes five units but is billed as one request. No effective date is published.
Multi-query requests drain the rate limit faster than the old per-request reading implied.
Current monitored baseline
HyperObserve currently monitors 4 scoped observations for Perplexity API. The baseline covers Agent API tier range, Search API throughput, Sonar model RPM, cumulative credit thresholds. Each value retains its plan, product, endpoint, region, account, or runtime qualifier so a future change can be compared against the correct scope.
| Constraint | Current value | Scope | Verified source |
|---|---|---|---|
| Agent API tier rangeQPS and RPM are independent limits and a request must fit both: Tier 1 is 3 QPS / 150 RPM, Tier 2 8 QPS / 500 RPM, Tier 3 17 QPS / 1,000 RPM. | Tier 0: 1 QPS / 50 RPM; Tier 4: 33 QPS / 4,000 RPM; Tier 5: 33 QPS / 8,000 RPM | Permanent cumulative-spend tiers | PerplexitySep 17, 2026 |
| Search API throughputEach query in a multi-query request consumes one unit while billing stays per request; the limit is a leaky bucket independent of usage tier. | 50 query units per second with a 50-unit burst | All usage tiers | PerplexitySep 17, 2026 |
| Sonar model RPMTier 0 allows 50 RPM for sonar, sonar-pro and sonar-reasoning-pro and 5 RPM for sonar-deep-research; Tier 5 allows 4,000 and 100 RPM; tiers unlock at $50, $250, $500, $1,000 and $5,000 cumulative spend. | Tier 0: 5–50 RPM; upper tiers: up to 100–4,000 RPM | Model and tier specific | PerplexitySep 17, 2026 |
| cumulative credit thresholdsOnce reached, tiers are retained and do not downgrade. | Tier 1 at $50; Tier 5 at $5,000 cumulative credits | API accounts | PerplexitySep 17, 2026 |
Tracked constraints and impact
Agent, Search, and Sonar API rate limits across permanent spend tiers. A change is recorded only when it alters a constraint developers can act on, such as capacity planning, request shaping, model selection, deployment configuration, storage design, retry behavior, or account budgeting.
Perplexity Agent API Rate Limits
Perplexity Agent API Rate Limits, verified against Perplexity API's official documentation with scope, implementation impact, caveats, and a direct check.
Perplexity Search API Rate Limit
Perplexity Search API Rate Limit, verified against Perplexity API's official documentation with scope, implementation impact, caveats, and a direct check.
Perplexity Sonar API Rate Limits
Perplexity Sonar API Rate Limits, verified against Perplexity API's official documentation with scope, implementation impact, caveats, and a direct check.
Perplexity API Usage Tier Thresholds
Perplexity API Usage Tier Thresholds, verified against Perplexity API's official documentation with scope, implementation impact, caveats, and a direct check.
How a change is verified
- Compare the current first-party statement with the previously stored observation, including the exact plan, model, runtime, endpoint, region, and account scope.
- Separate a true product change from documentation clarification, temporary capacity, configurable account state, or a limit that applies to a different execution path.
- Preserve the old and new values, official source, effective date when disclosed, detection date, and practical impact. If those elements cannot be supported, no historical event is published.
Official sources monitored
These first-party documents support the current Perplexity API baseline. HyperObserve links directly to them so you can confirm the live wording before making a production decision.
Agent QPS/RPM per tier, Search query units, Sonar per-model RPM, spend thresholds, 429 with Retry-After