Skip to content
GPTBUSTER GPTBUSTER GPTBUSTER

Benchmarks

This page provides performance metrics for the gptBuster HTTP API. While actual performance may vary depending on system resources (CPU, memory, etc.) and deployment configuration, the results below offer a useful baseline for estimating throughput and latency.

The general baseline performance is approximately 2,000 requests per second (RPS) across critical endpoints like /v1/challenge and /v1/verify, under typical production settings with rate-limiting and logging enabled. Higher throughput is possible with optimizations and under ideal conditions, as demonstrated in the benchmark results below.

Benchmarking Tips

To accurately benchmark your own deployment:

  • Disable rate-limiting: Turn off all rate limits and autopilot settings in your security group.
  • Disable request logging: Request logging can significantly impact performance. Disable it in the API Keys configuration, as it may reduce throughput by up to 50%.
  • Disable flood protection: Set the environment variable FLOOD_RATE_LIMIT to an empty value to disable the global flood-protection rate limit.

All benchmarks below were run locally (against localhost) on a MacBook Pro M3 Pro, with multi-threading enabled, against a single instance. Latency values reflect the full HTTP round-trip time. Tests conducted using Bombardier.

Benchmark Results

Challenge Generation

Endpoint: GET /v1/challenge

This endpoint generates a new GPTBUSTER challenge. It achieves approximately 5,200 requests per second, with an average latency of 1.90ms.

Statistics        Avg      Stdev        Max
  Reqs/sec      5257.57     409.02    5911.10
  Latency        1.90ms   496.87us    52.01ms
  HTTP codes:
    1xx - 0, 2xx - 105145, 3xx - 0, 4xx - 0, 5xx - 0
    others - 0
  Throughput:     5.29MB/s

Challenge Verification

Endpoint: POST /v1/verify

This endpoint verifies the solution to an GPTBUSTER challenge. It achieves approximately 6,300 requests per second, with an average latency of 1.58ms.

Statistics        Avg      Stdev        Max
  Reqs/sec      6321.77     537.82    7038.01
  Latency        1.58ms   496.79us    49.77ms
  HTTP codes:
    1xx - 0, 2xx - 126433, 3xx - 0, 4xx - 0, 5xx - 0
    others - 0
  Throughput:     7.99MB/s

Text Classification

Endpoint: POST /v1/classifier

This endpoint classifies 10KB of text using the built-in classifier model. It achieves approximately 930 requests per second, with an average latency of 10.73ms.

Statistics        Avg      Stdev        Max
  Reqs/sec       931.68     243.69    1219.46
  Latency       10.73ms    10.34ms   255.08ms
  HTTP codes:
    1xx - 0, 2xx - 18632, 3xx - 0, 4xx - 0, 5xx - 0
    others - 0
  Throughput:     6.03MB/s