I compared some of the framworks on TodoBackend project with some tweaks.

All these are backed by postgres. So obviously these benchmarks are not purely for the framework itself but the whole stack.

Phoenix

https://github.com/jeffweiss/todobackend-phoenix

wrk -t20 -c100 -d30S --timeout 2000  "http://localhost:3000/api/todos"
Running 30s test @ http://localhost:3000/api/todos
  20 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    23.76ms    3.25ms 114.35ms   96.11%
    Req/Sec   212.01     16.17   252.00     80.19%
  126864 requests in 30.09s, 82.27MB read
Requests/sec:   4216.04
Transfer/sec:      2.73MB

Sinatra

https://github.com/moredip/todo-backend-sinatra

removed the logging

wrk -t20 -c100 -d30S --timeout 2000  "http://localhost:3000/todos"
Running 30s test @ http://localhost:3000/todos
  20 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    13.63ms   11.83ms 173.18ms   88.35%
    Req/Sec   324.15    119.35   690.00     70.84%
  106624 requests in 30.10s, 93.75MB read
Requests/sec:   3542.85
Transfer/sec:      3.12MB

ExpressJs

https://github.com/dtao/todo-backend-express

Non Cluster mode

wrk -t20 -c100 -d30S --timeout 2000  "http://localhost:3000/"
Running 30s test @ http://localhost:3000/
  20 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    81.96ms   97.83ms 498.72ms   79.26%
    Req/Sec   144.97     93.23   540.00     63.28%
  81397 requests in 30.10s, 44.48MB read
Requests/sec:   2704.28
Transfer/sec:      1.48MB

Cluster mode

added cluster mode

wrk -t20 -c100 -d30S --timeout 2000  "http://localhost:3000/"
Running 30s test @ http://localhost:3000/
  20 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    21.58ms   20.51ms 408.22ms   90.05%
    Req/Sec   273.38     58.89   550.00     73.69%
  163169 requests in 30.07s, 89.16MB read
Requests/sec:   5425.57
Transfer/sec:      2.96MB

Elixir Maru

https://github.com/whitfieldc/maru_todo

wrk -t20 -c100 -d30S --timeout 2000  "http://localhost:3000/tasks"
Running 30s test @ http://localhost:3000/tasks
  20 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    21.77ms    5.43ms 143.73ms   95.62%
    Req/Sec   232.93     23.00   272.00     88.08%
  139189 requests in 30.07s, 68.25MB read
Requests/sec:   4628.15
Transfer/sec:      2.27MB

Vanilla Rails

https://github.com/hammerdr/todo-backend-rails

wrk -t20 -c100 -d30S --timeout 2000  "http://localhost:3000"
Running 30s test @ http://localhost:3000
  20 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    61.33ms   16.54ms 247.46ms   76.96%
    Req/Sec    60.09     32.24   141.00     60.18%
  26936 requests in 30.08s, 24.30MB read
Requests/sec:    895.37
Transfer/sec:    827.20KB

Rails 5

wrk -t20 -c100 -d30S --timeout 2000  "http://localhost:3000"
Running 30s test @ http://localhost:3000
  20 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    41.48ms   27.82ms 417.76ms   65.79%
    Req/Sec    90.61     63.23   280.00     63.91%
  29777 requests in 30.10s, 26.16MB read
Requests/sec:    989.23
Transfer/sec:      0.87MB