AI Speed Comparisons

Name: AIPerf Public LLM Latency Telemetry
License: https://creativecommons.org/licenses/by/4.0/

Head-to-head speed tests for the most-compared AI models — measured live from your own connection on time to first token (TTFT) and tokens per second (TPS).

ChatGPT (GPT-4o) vs Claude 3 Haiku
The two most-used assistants head to head — measured on time to first token and tokens per second.
Gemini 1.5 Flash vs ChatGPT (GPT-4o)
Google's fastest Flash tier against OpenAI's GPT-4o on real streaming latency.
Claude 3 Haiku vs Gemini 1.5 Flash
Two latency-optimized models compared on first-token wait and sustained throughput.
Llama 4 Scout vs ChatGPT (GPT-4o)
Groq-hosted Llama against GPT-4o — does specialized inference hardware win on speed?