We got about 770k req/s with the current version vs 582k from japronto using TechEmPower plaintext benchmarks and we still can get at least 30% or more performance once fully integrated with libuv + asyncio :D