### Answer There isn't an official "best" LLM designation from Retell AI—the right choice depends on your use case. Available models include:

gpt-4.1

,

gpt-4.1-mini

,

gpt-4.1-nano

,

gpt-5

,

gpt-5.1

,

gpt-5.2

,

gpt-5-mini

,

gpt-5-nano

,

claude-4.5-sonnet

,

claude-4.5-haiku

,

gemini-2.5-flash

,

gemini-2.5-flash-lite

[(1) ]() General guidance by use case: - Function calling / data collection: Use lower temperature (0.0–0.3) for reliability [(2) ]() - Customer service / sales: Balanced temperature (0.3–0.7) - Date/time calculations: GPT-4o or Sonnet tend to perform better than mini/Haiku models for accuracy [(3) ]() If cost isn't a concern and you want to minimize errors/hallucinations, larger models like

gpt-4.1

,

gpt-5

, or

claude-4.5-sonnet

are generally more capable. [(1) ]() You can also enable Fast Tier for more consistent latency on high-value calls (1.5x cost) . [(2) ]()

​ To learn more, see the sources I considered: [[1] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) [[2] Configure LLM Options](https://docs.retellai.com/build/llm-options) [[3] Discord message from @trxgames9630 on 2025-12-17](https://discord.com/channels/1186424376300679209/1294891368505937952/1450680465643016278) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

@AI Evy the reason I ask is that I am finding that gpt 5 isn't preforming as well as gpt 4x

### Answer --- Your experience aligns with some known observations about GPT-5 on Retell AI: Known GPT-5 limitations: - GPT-5 does not support

NO_RESPONSE_NEEDED

— it reads it aloud instead of using it to suppress a response [(1) ]() - Some users report GPT-5 agents getting stuck in loops, repeating the same question even after the user answers [(2) ]() - Testing shows GPT-5 models can be verbose in replies [(3) ]() - GPT-5 has higher latency (~1000ms) compared to GPT-4.1 (~720ms) [(4) ]() Recommendations from Retell AI: - For most real-time voice agent use cases where speed is critical, GPT-5 Mini or GPT-4.1 are better choices - GPT-5 becomes more viable if future improvements reduce its latency - You can be more assertive with prompts (e.g., "don't repeat the question unless asked to") to mitigate some behaviors [(2) ]() - If GPT-4.x works better for your use case, staying with it is a valid approach GPT-4.1 advantages: - Lower latency (720ms vs 1000ms for GPT-5) [(4) ]() - Better instruction-following and system prompt compliance [(5) ]() - Optimized for real-time conversation flow If you're seeing better results with GPT-4.x, that's expected for many real-time voice scenarios.

​ To learn more, see the sources I considered: [[1] GPT-5](https://discord.com/channels/1186424376300679209/1436024080527790211) [[2] Issue with GPT-5 model.](https://discord.com/channels/1186424376300679209/1411997452315263059) [[3] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) [[4] Retell AI Launches GPT-5 for Voice Agents – Higher Intelligence, Smarter Conversations | Retell AI](https://www.retellai.com/blog/gpt-5-now-on-retell-smarter-ai-voice-agents-with-reasoning-power) [[5] Smarter AI Phone Calls with ChatGPT-4.1 | Retell AI | Retell AI](https://www.retellai.com/blog/chatgpt-4-1-ai-phone-calling-retell) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer ### Answer Based on community feedback, here are some voices users have mentioned: Mentioned as natural-sounding: - Marissa — recommended by a user when asked about natural-sounding voices [(1) ]() - *Emily (11labs-Emily) * — described as "great" by a user exploring voice options [(2) ]() Mixed feedback: - Chloe — was previously considered "upbeat and positive" for customer service, but some users report it now sounds "lower energy" and "depressed," with unexpected speed changes [(3) ]() - *Jennifer (ElevenLabs) * — described as "good, but not great" by a user seeking the best female voice [(4) ]() General guidance: Retell AI offers pre-trained voices optimized for conversational depth, and you can also integrate custom voices from providers like ElevenLabs. [(5) ]() Voice quality can be subjective and may depend on your use case (customer service vs. sales vs. support) . Testing a few options with your specific prompts is recommended.

​ To learn more, see the sources I considered: [[1] Discord message from @kingstonelunaroid on 2025-06-19](https://discord.com/channels/1186424376300679209/1294891368505937952/1385320098893533316) [[2] Voice ID](https://discord.com/channels/1186424376300679209/1378571804330823741) [[3] "Chloe" voice sounds different as of a few days ago](https://discord.com/channels/1186424376300679209/1271547891533615278) [[4] Discord message from @fiery_moose_79379 on 2025-07-10](https://discord.com/channels/1186424376300679209/1294891368505937952/1392945981775151254) [[5] Top AI Voice Trends Reshaping Call Centers in 2025 | Retell AI | Retell AI](https://www.retellai.com/blog/ai-voice-trends-enterprise-call-centers-2025) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.