```
======================================================================
Milkha Cross-Language Benchmark: FastAPI vs Mojo
======================================================================

Running FastAPI benchmarks (10000 iterations)...
----------------------------------------------------------------------
Endpoint                          Latency           Throughput
----------------------------------------------------------------------
  GET /                        806.131 us/req         1240 req/s
  GET /users/{id}              748.120 us/req         1337 req/s
  POST /items                  779.470 us/req         1283 req/s
  DELETE /users/{id}           810.901 us/req         1233 req/s
----------------------------------------------------------------------

Expected Milkha (Mojo) performance:
  FastAPI uses Python interpreter + async overhead per request.
  Milkha compiles to native code via Mojo, eliminating:
    - Python interpreter overhead
    - Per-request async machinery
    - GIL contention
  Expected speedup: ~10-50x for simple endpoints (based on
  FastAPi vs compiled-language benchmarks in TFB).

  Example projections:
  Endpoint                          FastAPI        Expected Mojo
  ------------------------- --------------- --------------------
  GET /                           1240 req/s           37215 req/s
  GET /users/{id}                 1337 req/s           40101 req/s
  POST /items                     1283 req/s           38488 req/s
  DELETE /users/{id}              1233 req/s           36996 req/s

======================================================================
Note: Actual Mojo performance depends on hardware, Flare version,
and compiler optimizations. Run with `mojo build -D ASSERT=none -O3`
for production performance.
======================================================================
```
