GraphQL Federation Benchmark
An open benchmark of GraphQL federation routers/gateways. Each router is evaluated under the same conditions, on the same machine with the same subgraphs and the same GraphQL operation, so you can compare throughput, latency, memory and CPU on equal terms.
New results are published every day from a public workflow that anyone can inspect or rerun on their own hardware.
Latest run
- Run
- #402
- Completed
Results at a glance
Fusion is the fastest router in every scenario of run 402.
- Real-world
- 1,947 requests/s
- 4% faster than Hive Router
- Synthetic
- 2,653 requests/s
- 5.2% faster than Hive Router
- Burst traffic
- 2,733 requests/s
- 18.8% faster than Hive Router
Burst traffic results
GraphQL Federation specification
RouterMedian requests/sHTTP requests per second, as the median across the measured runs. In this scenario it is the average over the whole run, not the rate at 500 virtual users.
2,733, highest reported
All metrics for Fusion Nightly
- p95 latency95% of requests completed within this time.
- 184 ms
- p99.9 latency99.9% of requests completed within this time.
- 257 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 526 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 97.20%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.6%
2,694
All metrics for Fusion Nightly
- p95 latency95% of requests completed within this time.
- 204 ms
- p99.9 latency99.9% of requests completed within this time.
- 624 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 617 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 97.70%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.3%
2,448
All metrics for Fusion
- p95 latency95% of requests completed within this time.
- 206 ms
- p99.9 latency99.9% of requests completed within this time.
- 675 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 598 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 98.10%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.6%
833
All metrics for feddi
- p95 latency95% of requests completed within this time.
- 1,050 ms
- p99.9 latency99.9% of requests completed within this time.
- 2,244 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 1,454 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 80.60%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±2.9%
| Router/gateway | Median requests/sHTTP requests per second, as the median across the measured runs. In this scenario it is the average over the whole run, not the rate at 500 virtual users. | p95 latency95% of requests completed within this time. | p99.9 latency99.9% of requests completed within this time. | Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes. | Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads. | ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure. | Run varianceHow much throughput varied across the measured runs, as the coefficient of variation. |
|---|---|---|---|---|---|---|---|
2,733, highest reported | 184 ms | 257 ms | 526 MiB | 97.20 % | 100% | ±1.6% | |
2,694 | 204 ms | 624 ms | 617 MiB | 97.70 % | 100% | ±1.3% | |
2,448 | 206 ms | 675 ms | 598 MiB | 98.10 % | 100% | ±1.6% | |
833 | 1,050 ms | 2,244 ms | 1,454 MiB | 80.60 % | 100% | ±2.9% |
Apollo Federation v2
RouterMedian requests/sHTTP requests per second, as the median across the measured runs. In this scenario it is the average over the whole run, not the rate at 500 virtual users.
2,386, highest reported
All metrics for Fusion Nightly
- p95 latency95% of requests completed within this time.
- 213 ms
- p99.9 latency99.9% of requests completed within this time.
- 682 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 622 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 97.80%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±2.1%
2,300
All metrics for Hive Router
- p95 latency95% of requests completed within this time.
- 210 ms
- p99.9 latency99.9% of requests completed within this time.
- 258 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 198 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 97.70%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±2.7%
1,952
All metrics for Grafbase Gateway
- p95 latency95% of requests completed within this time.
- 164 ms
- p99.9 latency99.9% of requests completed within this time.
- 10,286 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 193 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 95.70%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 65.98%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±7.4%
1,295
All metrics for Cosmo Router
- p95 latency95% of requests completed within this time.
- 357 ms
- p99.9 latency99.9% of requests completed within this time.
- 423 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 567 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 98.70%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.5%
1,277
All metrics for Caliban
- p95 latency95% of requests completed within this time.
- 437 ms
- p99.9 latency99.9% of requests completed within this time.
- 763 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 10,925 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 99.40%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.6%
471
All metrics for Hive Gateway · Router Runtime
- p95 latency95% of requests completed within this time.
- 1,000 ms
- p99.9 latency99.9% of requests completed within this time.
- 1,531 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 3,185 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 99.70%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.0%
399
All metrics for Apollo Router
- p95 latency95% of requests completed within this time.
- 1,021 ms
- p99.9 latency99.9% of requests completed within this time.
- 9,692 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 504 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 99.50%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 81.42%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±2.5%
361
All metrics for Apollo Router
- p95 latency95% of requests completed within this time.
- 846 ms
- p99.9 latency99.9% of requests completed within this time.
- 21,561 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 483 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 99.80%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 96.97%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.6%
244
All metrics for Hive Gateway
- p95 latency95% of requests completed within this time.
- 2,086 ms
- p99.9 latency99.9% of requests completed within this time.
- 4,085 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 3,935 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 99.20%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.0%
231
All metrics for Apollo Gateway
- p95 latency95% of requests completed within this time.
- 1,974 ms
- p99.9 latency99.9% of requests completed within this time.
- 2,922 ms
- Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes.
- 2,461 MiB
- Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads.
- 99.30%
- ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure.
- 100%
- Run varianceHow much throughput varied across the measured runs, as the coefficient of variation.
- ±1.0%
| Router/gateway | Median requests/sHTTP requests per second, as the median across the measured runs. In this scenario it is the average over the whole run, not the rate at 500 virtual users. | p95 latency95% of requests completed within this time. | p99.9 latency99.9% of requests completed within this time. | Peak memoryThe highest sampled resident memory (RSS) of the router process and its child processes. | Average CPUAverage utilization of the router’s 4 hardware threads. 100% means both of its cores are fully busy on both threads. | ReliabilityThe share of requests that passed the checks for HTTP status, GraphQL errors and response structure. | Run varianceHow much throughput varied across the measured runs, as the coefficient of variation. |
|---|---|---|---|---|---|---|---|
2,386, highest reported | 213 ms | 682 ms | 622 MiB | 97.80 % | 100% | ±2.1% | |
2,300 | 210 ms | 258 ms | 198 MiB | 97.70 % | 100% | ±2.7% | |
1,952 | 164 ms | 10,286 ms | 193 MiB | 95.70 % | 65.98% | ±7.4% | |
1,295 | 357 ms | 423 ms | 567 MiB | 98.70 % | 100% | ±1.5% | |
1,277 | 437 ms | 763 ms | 10,925 MiB | 99.40 % | 100% | ±1.6% | |
471 | 1,000 ms | 1,531 ms | 3,185 MiB | 99.70 % | 100% | ±1.0% | |
399 | 1,021 ms | 9,692 ms | 504 MiB | 99.50 % | 81.42% | ±2.5% | |
361 | 846 ms | 21,561 ms | 483 MiB | 99.80 % | 96.97% | ±1.6% | |
244 | 2,086 ms | 4,085 ms | 3,935 MiB | 99.20 % | 100% | ±1.0% | |
231 | 1,974 ms | 2,922 ms | 2,461 MiB | 99.30 % | 100% | ±1.0% |
Throughput over time
Median requests per second by date.
requests/s
22 Jul
GraphQL Federation specification
Apollo Federation v2
| Router | Run dates, versions and median requests per second |
|---|---|
| Apollo Gateway (Node 24) | 2026-07-22T11:52:09.000Z: 2.14.2, 233 requests/s, 2026-07-27T16:03:23.000Z: 2.14.3, 238 requests/s, 2026-07-30T16:01:56.000Z: 2.14.3, 228 requests/s, 2026-08-01T16:07:23.000Z: 2.14.3, 236 requests/s, 2026-08-05T16:56:21.000Z: 2.14.3, 236 requests/s, 2026-08-11T11:47:47.000Z: 2.14.3, 229 requests/s, 2026-08-20T16:40:29.000Z: 2.14.3, 233 requests/s, 2026-08-24T16:35:28.000Z: 2.14.3, 236 requests/s, 2026-08-27T17:26:23.000Z: 2.14.3, 236 requests/s, 2026-08-29T17:35:51.000Z: 2.14.3, 235 requests/s, 2026-08-31T16:39:02.000Z: 2.14.3, 236 requests/s, 2026-09-02T16:40:25.000Z: 2.14.3, 232 requests/s, 2026-09-09T16:36:14.000Z: 2.14.3, 235 requests/s, 2026-09-13T16:41:16.000Z: 2.14.3, 231 requests/s, 2026-09-17T16:37:09.000Z: 2.14.3, 229 requests/s, 2026-09-19T16:35:00.000Z: 2.14.3, 234 requests/s, 2026-09-22T16:40:03.000Z: 2.14.3, 237 requests/s, 2026-10-17T10:49:56.000Z: 2.14.4, 231 requests/s |
| Apollo Router 2.x (Rust) | 2026-07-22T11:52:09.000Z: 2.16.0, 406 requests/s, 2026-07-27T16:03:23.000Z: 2.16.1, 361 requests/s, 2026-07-30T16:01:56.000Z: 2.16.1, 362 requests/s, 2026-08-01T16:07:23.000Z: 2.16.1, 359 requests/s, 2026-08-05T16:56:21.000Z: 2.16.1, 402 requests/s, 2026-08-11T11:47:47.000Z: 2.16.1, 405 requests/s, 2026-08-20T16:40:29.000Z: 2.16.1, 389 requests/s, 2026-08-24T16:35:28.000Z: 2.16.1, 395 requests/s, 2026-08-27T17:26:23.000Z: 2.16.1, 405 requests/s, 2026-08-29T17:35:51.000Z: 2.16.1, 401 requests/s, 2026-08-31T16:39:02.000Z: 2.16.1, 404 requests/s, 2026-09-02T16:40:25.000Z: 2.16.1, 395 requests/s, 2026-09-09T16:36:14.000Z: 2.16.1, 367 requests/s, 2026-09-13T16:41:16.000Z: 2.16.1, 365 requests/s, 2026-09-17T16:37:09.000Z: 2.16.1, 403 requests/s, 2026-09-19T16:35:00.000Z: 2.16.1, 401 requests/s, 2026-09-22T16:40:03.000Z: 2.16.1, 399 requests/s, 2026-10-17T10:49:56.000Z: 2.18.0, 399 requests/s |
| Apollo Router 3.x (Rust) | 2026-10-16T10:24:10.000Z: 3.0.0-preview.0, 360 requests/s, 2026-10-17T10:49:56.000Z: 3.0.0-preview.0, 361 requests/s |
| Caliban (JVM 25) | 2026-10-16T10:24:10.000Z: 3.2.0, 1,267 requests/s, 2026-10-17T10:49:56.000Z: 3.2.0, 1,277 requests/s |
| Cosmo Router (Go) | 2026-07-22T11:52:09.000Z: 0.331.1, 1,168 requests/s, 2026-07-27T16:03:23.000Z: 0.334.0, 1,124 requests/s, 2026-07-30T16:01:56.000Z: 0.334.0, 1,173 requests/s, 2026-08-01T16:07:23.000Z: 0.334.0, 1,153 requests/s, 2026-08-05T16:56:21.000Z: 0.334.0, 1,115 requests/s, 2026-08-11T11:47:47.000Z: 0.334.0, 1,169 requests/s, 2026-08-20T16:40:29.000Z: 0.334.0, 1,115 requests/s, 2026-08-24T16:35:28.000Z: 0.334.0, 1,123 requests/s, 2026-08-27T17:26:23.000Z: 0.334.0, 1,112 requests/s, 2026-08-29T17:35:51.000Z: 0.334.0, 1,125 requests/s, 2026-08-31T16:39:02.000Z: 0.334.0, 1,155 requests/s, 2026-09-02T16:40:25.000Z: 0.334.0, 1,134 requests/s, 2026-09-09T16:36:14.000Z: 0.334.0, 1,163 requests/s, 2026-09-13T16:41:16.000Z: 0.334.0, 1,165 requests/s, 2026-09-17T16:37:09.000Z: 0.334.0, 1,135 requests/s, 2026-09-19T16:35:00.000Z: 0.334.0, 1,157 requests/s, 2026-09-22T16:40:03.000Z: 0.334.0, 1,150 requests/s, 2026-10-17T10:49:56.000Z: 0.359.0, 1,295 requests/s |
| Grafbase Gateway (Rust) | 2026-07-22T11:52:09.000Z: 0.53.5, 1,980 requests/s, 2026-07-27T16:03:23.000Z: 0.53.5, 1,766 requests/s, 2026-07-30T16:01:56.000Z: 0.53.5, 1,892 requests/s, 2026-08-01T16:07:23.000Z: 0.53.5, 1,820 requests/s, 2026-08-05T16:56:21.000Z: 0.53.5, 1,799 requests/s, 2026-08-11T11:47:47.000Z: 0.53.5, 1,914 requests/s, 2026-08-20T16:40:29.000Z: 0.53.5, 1,921 requests/s, 2026-08-24T16:35:28.000Z: 0.53.5, 1,683 requests/s, 2026-08-27T17:26:23.000Z: 0.53.5, 1,850 requests/s, 2026-08-29T17:35:51.000Z: 0.53.5, 1,992 requests/s, 2026-08-31T16:39:02.000Z: 0.53.5, 1,870 requests/s, 2026-09-02T16:40:25.000Z: 0.53.5, 1,679 requests/s, 2026-09-09T16:36:14.000Z: 0.53.5, 1,721 requests/s, 2026-09-13T16:41:16.000Z: 0.53.5, 1,986 requests/s, 2026-09-17T16:37:09.000Z: 0.53.5, 1,885 requests/s, 2026-09-19T16:35:00.000Z: 0.53.5, 1,964 requests/s, 2026-09-22T16:40:03.000Z: 0.53.5, 2,047 requests/s, 2026-10-17T10:49:56.000Z: 0.53.5, 1,952 requests/s |
| Hive Gateway (Node 24) | 2026-07-22T11:52:09.000Z: 2.10.4, 247 requests/s, 2026-07-27T16:03:23.000Z: 2.10.8, 246 requests/s, 2026-07-30T16:01:56.000Z: 2.10.8, 248 requests/s, 2026-08-01T16:07:23.000Z: 2.10.8, 242 requests/s, 2026-08-05T16:56:21.000Z: 2.10.8, 248 requests/s, 2026-08-11T11:47:47.000Z: 2.10.8, 240 requests/s, 2026-08-20T16:40:29.000Z: 2.10.8, 238 requests/s, 2026-08-24T16:35:28.000Z: 2.10.8, 242 requests/s, 2026-08-27T17:26:23.000Z: 2.10.8, 237 requests/s, 2026-08-29T17:35:51.000Z: 2.10.8, 250 requests/s, 2026-08-31T16:39:02.000Z: 2.10.8, 245 requests/s, 2026-09-02T16:40:25.000Z: 2.10.8, 243 requests/s, 2026-09-09T16:36:14.000Z: 2.10.8, 249 requests/s, 2026-09-13T16:41:16.000Z: 2.10.8, 238 requests/s, 2026-09-17T16:37:09.000Z: 2.10.8, 240 requests/s, 2026-09-19T16:35:00.000Z: 2.10.8, 244 requests/s, 2026-09-22T16:40:03.000Z: 2.10.8, 246 requests/s, 2026-10-17T10:49:56.000Z: 2.15.2, 244 requests/s |
| Hive Gateway · Router Runtime (Node 24) | 2026-07-22T11:52:09.000Z: 2.10.4, 541 requests/s, 2026-07-27T16:03:23.000Z: 2.10.8, 529 requests/s, 2026-07-30T16:01:56.000Z: 2.10.8, 516 requests/s, 2026-08-01T16:07:23.000Z: 2.10.8, 539 requests/s, 2026-08-05T16:56:21.000Z: 2.10.8, 540 requests/s, 2026-08-11T11:47:47.000Z: 2.10.8, 518 requests/s, 2026-08-20T16:40:29.000Z: 2.10.8, 549 requests/s, 2026-08-24T16:35:28.000Z: 2.10.8, 522 requests/s, 2026-08-27T17:26:23.000Z: 2.10.8, 531 requests/s, 2026-08-29T17:35:51.000Z: 2.10.8, 533 requests/s, 2026-08-31T16:39:02.000Z: 2.10.8, 545 requests/s, 2026-09-02T16:40:25.000Z: 2.10.8, 543 requests/s, 2026-09-09T16:36:14.000Z: 2.10.8, 527 requests/s, 2026-09-13T16:41:16.000Z: 2.10.8, 526 requests/s, 2026-09-17T16:37:09.000Z: 2.10.8, 523 requests/s, 2026-09-19T16:35:00.000Z: 2.10.8, 543 requests/s, 2026-09-22T16:40:03.000Z: 2.10.8, 536 requests/s, 2026-10-17T10:49:56.000Z: 2.15.2, 471 requests/s |
| Hive Router (Rust) | 2026-07-22T11:52:09.000Z: 0.0.83, 2,204 requests/s, 2026-07-27T16:03:23.000Z: 0.0.84, 2,236 requests/s, 2026-07-30T16:01:56.000Z: 0.0.84, 2,191 requests/s, 2026-08-01T16:07:23.000Z: 0.0.84, 2,386 requests/s, 2026-08-05T16:56:21.000Z: 0.0.84, 2,198 requests/s, 2026-08-11T11:47:47.000Z: 0.0.84, 2,257 requests/s, 2026-08-20T16:40:29.000Z: 0.0.84, 2,282 requests/s, 2026-08-24T16:35:28.000Z: 0.0.84, 2,294 requests/s, 2026-08-27T17:26:23.000Z: 0.0.84, 2,216 requests/s, 2026-08-29T17:35:51.000Z: 0.0.84, 2,385 requests/s, 2026-08-31T16:39:02.000Z: 0.0.84, 2,302 requests/s, 2026-09-02T16:40:25.000Z: 0.0.84, 2,365 requests/s, 2026-09-09T16:36:14.000Z: 0.0.84, 2,390 requests/s, 2026-09-13T16:41:16.000Z: 0.0.84, 2,380 requests/s, 2026-09-17T16:37:09.000Z: 0.0.84, 2,223 requests/s, 2026-09-19T16:35:00.000Z: 0.0.84, 2,199 requests/s, 2026-09-22T16:40:03.000Z: 0.0.84, 2,207 requests/s, 2026-10-17T10:49:56.000Z: 0.3.5, 2,300 requests/s |
| Fusion (.NET 10) | 2026-07-22T11:52:09.000Z: 16.5.1, 2,355 requests/s, 2026-07-27T16:03:23.000Z: 16.5.1, 2,351 requests/s, 2026-07-30T16:01:56.000Z: 16.5.1, 2,356 requests/s, 2026-08-01T16:07:23.000Z: 16.5.1, 2,269 requests/s, 2026-08-05T16:56:21.000Z: 16.5.1, 2,285 requests/s, 2026-08-11T11:47:47.000Z: 16.6.0, 2,413 requests/s, 2026-08-20T16:40:29.000Z: 16.6.1, 2,325 requests/s, 2026-08-24T16:35:28.000Z: 16.6.1, 2,303 requests/s, 2026-08-27T17:26:23.000Z: 16.6.1, 2,397 requests/s, 2026-08-29T17:35:51.000Z: 16.6.2, 2,398 requests/s, 2026-08-31T16:39:02.000Z: 16.6.2, 2,324 requests/s, 2026-09-02T16:40:25.000Z: 16.6.2, 2,406 requests/s, 2026-09-09T16:36:14.000Z: 16.6.4, 2,323 requests/s, 2026-09-13T16:41:16.000Z: 16.6.6, 2,326 requests/s, 2026-09-17T16:37:09.000Z: 16.6.6, 2,433 requests/s, 2026-09-19T16:35:00.000Z: 16.6.6, 2,412 requests/s, 2026-09-22T16:40:03.000Z: 16.6.6, 2,427 requests/s, 2026-10-17T10:49:56.000Z: 16.6.8, 2,448 requests/s |
| Fusion Nightly (.NET 10) | 2026-07-26T16:01:42.000Z: 16.6.0-p.3, 2,320 requests/s, 2026-07-29T16:13:06.000Z: 16.6.0-p.6, 2,396 requests/s, 2026-07-31T16:10:01.000Z: 16.6.0-p.8, 2,351 requests/s, 2026-08-03T16:01:43.000Z: 16.6.0-p.13, 2,393 requests/s, 2026-08-14T16:49:42.000Z: 16.6.1-p.2, 2,319 requests/s, 2026-08-23T16:36:55.000Z: 16.6.2-p.6, 2,327 requests/s, 2026-08-25T16:40:02.000Z: 16.6.2-p.7, 2,351 requests/s, 2026-08-28T19:36:23.000Z: 16.6.2-p.8, 2,377 requests/s, 2026-08-30T16:37:14.000Z: 16.7.0-p.1, 2,328 requests/s, 2026-08-31T16:39:02.000Z: 16.7.0-p.1, 2,428 requests/s, 2026-09-02T16:40:25.000Z: 16.7.0-p.2, 2,403 requests/s, 2026-09-09T16:36:14.000Z: 16.7.0-p.5, 2,298 requests/s, 2026-09-13T16:41:16.000Z: 16.7.0-p.6, 2,368 requests/s, 2026-09-17T16:37:09.000Z: 16.7.0-p.7, 2,314 requests/s, 2026-09-19T16:35:00.000Z: 16.7.0-p.8, 2,295 requests/s, 2026-09-22T16:40:03.000Z: 16.7.0-p.9, 2,320 requests/s, 2026-10-17T10:49:56.000Z: 16.7.0-p.28, 2,694 requests/s |
| Fusion Nightly (.NET 11) | 2026-07-26T16:01:42.000Z: 16.6.0-p.3, 2,460 requests/s, 2026-07-29T16:13:06.000Z: 16.6.0-p.6, 2,450 requests/s, 2026-07-31T16:10:01.000Z: 16.6.0-p.8, 2,468 requests/s, 2026-08-03T16:01:43.000Z: 16.6.0-p.13, 2,367 requests/s, 2026-08-14T16:49:42.000Z: 16.6.1-p.2, 2,481 requests/s, 2026-08-23T16:36:55.000Z: 16.6.2-p.6, 2,383 requests/s, 2026-08-25T16:40:02.000Z: 16.6.2-p.7, 2,385 requests/s, 2026-08-28T19:36:23.000Z: 16.6.2-p.8, 2,498 requests/s, 2026-08-30T16:37:14.000Z: 16.7.0-p.1, 2,492 requests/s, 2026-08-31T16:39:02.000Z: 16.7.0-p.1, 2,404 requests/s, 2026-09-02T16:40:25.000Z: 16.7.0-p.2, 2,485 requests/s, 2026-09-09T16:36:14.000Z: 16.7.0-p.5, 2,369 requests/s, 2026-09-13T16:41:16.000Z: 16.7.0-p.6, 2,359 requests/s, 2026-09-17T16:37:09.000Z: 16.7.0-p.7, 2,344 requests/s, 2026-09-19T16:35:00.000Z: 16.7.0-p.8, 2,474 requests/s, 2026-09-22T16:40:03.000Z: 16.7.0-p.9, 2,464 requests/s, 2026-10-17T10:49:56.000Z: 16.7.0-p.28, 2,733 requests/s |
| Fusion Nightly (.NET 10) | 2026-07-26T16:01:42.000Z: 16.6.0-p.3, 2,215 requests/s, 2026-07-29T16:13:06.000Z: 16.6.0-p.6, 2,196 requests/s, 2026-07-31T16:10:01.000Z: 16.6.0-p.8, 2,094 requests/s, 2026-08-03T16:01:43.000Z: 16.6.0-p.13, 2,207 requests/s, 2026-08-14T16:49:42.000Z: 16.6.1-p.2, 2,129 requests/s, 2026-08-23T16:36:55.000Z: 16.6.2-p.6, 2,159 requests/s, 2026-08-25T16:40:02.000Z: 16.6.2-p.7, 2,245 requests/s, 2026-08-28T19:36:23.000Z: 16.6.2-p.8, 2,237 requests/s, 2026-08-30T16:37:14.000Z: 16.7.0-p.1, 2,231 requests/s, 2026-08-31T16:39:02.000Z: 16.7.0-p.1, 2,238 requests/s, 2026-09-02T16:40:25.000Z: 16.7.0-p.2, 2,136 requests/s, 2026-09-09T16:36:14.000Z: 16.7.0-p.5, 2,223 requests/s, 2026-09-13T16:41:16.000Z: 16.7.0-p.6, 2,234 requests/s, 2026-09-17T16:37:09.000Z: 16.7.0-p.7, 2,235 requests/s, 2026-09-19T16:35:00.000Z: 16.7.0-p.8, 2,258 requests/s, 2026-09-22T16:40:03.000Z: 16.7.0-p.9, 2,186 requests/s, 2026-10-17T10:49:56.000Z: 16.7.0-p.28, 2,386 requests/s |
| feddi (JVM 25) | 2026-10-17T10:49:56.000Z: e91d605ed222, 833 requests/s |
Test conditions
Procedure
Every scenario runs these steps in order, one router at a time.
Start
The router runs on 2 dedicated physical cores (4 hardware threads). CPU boost is disabled to reduce clock-speed variation.
Warm up
One run of the full scenario length primes the router before anything is measured. Its results are discarded.
Measure
9 more runs of the same length follow. Each one sends the same GraphQL operation with the load profile of the scenario.
Check responses
A request only counts as a success when it returns HTTP 200 without GraphQL errors.
Report the median
Results use the median requests per second of the 9 runs, and run variance shows how much the runs differed.
Hardware
- Processor
- AMD Ryzen 7 5700G with Radeon Graphics
- CPU topology
- 8 physical cores (16 hardware threads, SMT).
- Memory
- 122 GB
Software
- Operating system
- Linux 6.8.0-147-generic
- Subgraphs
- Hot Chocolate 16.6.8 (.NET 10.0.13)
CPU pinning
- Hardware thread 0, physical core 0: OS & monitoring
- Hardware thread 8, physical core 0: OS & monitoring
- Hardware thread 1, physical core 1: Load generator
- Hardware thread 9, physical core 1: Load generator
- Hardware thread 2, physical core 2: Router
- Hardware thread 10, physical core 2: Router
- Hardware thread 3, physical core 3: Router
- Hardware thread 11, physical core 3: Router
- Hardware thread 4, physical core 4: Subgraphs
- Hardware thread 12, physical core 4: Subgraphs
- Hardware thread 5, physical core 5: Subgraphs
- Hardware thread 13, physical core 5: Subgraphs
- Hardware thread 6, physical core 6: Subgraphs
- Hardware thread 14, physical core 6: Subgraphs
- Hardware thread 7, physical core 7: Subgraphs
- Hardware thread 15, physical core 7: Subgraphs
- OS & monitoring
- Load generator
- Router
- Subgraphs
Processes are pinned with taskset; all other processes are moved onto core 0. Kernel threads and interrupts can still run on any core.
Published by ChilliCream, the team behind Fusion, using a harness based on The Guild's gateway benchmarks.