Llama-3-70B instruct benchmarks

I find it very suprising, in my testing @FireworksAI_HQ is almost as fast as @GroqInc!

Time taken for get_groq_response_requests: 1.28 seconds

Time taken for get_together_ai_response_requests: 2.60 seconds

Time taken for get_fireworks_ai_response_requests: 1.42 seconds

Groq is still faster 1.28s vs 1.42s for fireworks, but I doubt they build their own chip