Hi, tried moving from gpt-4.1 to gpt- 5.4-mini with my Vapi agents, seeing noticable average reasoning latency increase like almost 2x Is this expected? Any insights? Trying to reduce our llm costs

Hey Alcarine, Thanks for reaching out. the increase you’re seeing is expected because gpt-5.4-mini is a reasoning model, while gpt-4.1 is non-reasoning and typically lower latency for simple conversational flows I'd suggest: - gpt-4.1-mini or gpt-4.1-nano for better cost/latency tradeoffs - gpt-realtime if you want the lowest end to end voice latency here are all the doc ref: - https://docs.vapi.ai/whats-new/2026/3/31 - https://docs.vapi.ai/whats-new/2025/4/15 - https://docs.vapi.ai/openai-realtime Kind Regards, Akash

Hey @Alcarine . I don't work for Vapi, but I suggest you using 5.1 or 4.1. I prefer 5.1, I noticed these two being awesome for voice convo ai! 🙂

hi, thank you for suggestions and insights !

Glad to hear that you found this helpful. Please feel free to reach out if you ever encounter any issues or have any questions. Kind regards, Akash