See here for our TTFT metric benchmarks: https://artificialanalysis.ai/models/llama-3-1-instruct-70b/...