https://retellai.com/?utm_source=discord&utm_medium=seo logo
choosing multilingual model makes llm latency bigg...
# ❓・ai-evy-support
d
@AI Evy choosing multilingual model makes llm latency bigger?
a
Thanks for tagging me @dropsofsteel. I'll look through relevant sources and get back to you shortly.
### Answer The available sources don't explicitly state whether enabling multilingual or auto-language-detection features increases LLM latency in Retell AI. What the documentation does mention: - Features marked with a turtle icon 🐢 in your configuration indicate they may add latency. Check if your estimated latency is higher than 1.5s and consider disabling some features to reduce it. [(1) ]() - Normal LLM latency range is 500ms–900ms. If consistently above 900ms, you can use the "fast tier" or switch LLM providers. I'd recommend checking your agent configuration for the turtle icon next to the multilingual setting—that would confirm if it's flagged as a latency-adding feature. Would you like a general answer based on how multilingual processing typically affects latency?
​ To learn more, see the sources I considered: [[1] Troubleshoot high latency](https://docs.retellai.com/reliability/troubleshoot-latency) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
d
@AI Evy what is the "fast tier"?
a
Thanks for the follow-up, @dropsofsteel. I'll get back to you soon.
### Answer --- Fast Tier is a premium performance option for OpenAI models that routes your LLM calls through dedicated, high-priority infrastructure. [(1) ]() Key benefits: - Predictable latency – consistent response times for every call , [(2) ]() - Higher availability – priority access to compute resources [(1) ]() - Reduced variance – minimal fluctuation in processing speeds - Uncapped scale [(2) ]() Performance benchmarks: - 50% reduction in latency variance [(1) ]() - 25% improvement in average response time - 99.9% availability vs 99.5% standard Pricing: 1.5x the standard model cost (reduced from the previous 2x rate) . , [(2) ]() When to use it: - High-value customer interactions - Time-sensitive operations - Premium service tiers - Demos and sales calls [(1) ]() May not be necessary for: - Internal testing - Low-volume or non-critical calls - Cost-sensitive applications
​ To learn more, see the sources I considered: [[1] Configure LLM Options](https://docs.retellai.com/build/llm-options) [[2] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
d
@AI Evy how to activate Fast Tier ?
a
Thanks for the follow-up, @dropsofsteel. I'll get back to you soon.
### Answer The available sources describe what Fast Tier is and its benefits, but they don't provide step-by-step instructions on how to activate it in the dashboard or via API. [(1) ]() Based on the documentation, Fast Tier is available for OpenAI models and is configured as an LLM option. , [(2) ]() To find the activation setting, I'd recommend: - Navigating to your agent's LLM configuration in the Retell dashboard - Looking for a Fast Tier toggle or option under the LLM settings For detailed activation steps, check out: [Configure LLM Options]() [(1) ]() Would you like me to provide a general answer on where this setting is typically located?
​ To learn more, see the sources I considered: [[1] Configure LLM Options](https://docs.retellai.com/build/llm-options) [[2] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
d
@AI Evy зачем нужна такая штука как "metadata": { "user_phone_number": String($('Loop Over Items').item.json['Phone Number']) },
a
Thanks for the follow-up, @dropsofsteel. I'll get back to you soon.
### Answer ### Ответ Metadata — это произвольный объект для хранения пользовательских данных, связанных со звонком. [(1) ]() Основные цели использования: 1. Хранение идентификаторов — можно сохранить внутренний ID клиента, контакт ID из CRM (Salesforce, HubSpot и т.д.) для последующей обработки , [(2) ](

▾

) 2. Отслеживание и поиск — metadata позволяет связать звонок с конкретной записью в вашей системе. Например, в видео показано, как
user_phone
используется для поиска контакта в Google Sheets после звонка 3. Доступ через API и вебхуки — metadata доступна через Get Call API и включается в webhook payloads [(3) ]() Пример из документации:
Copy code
javascript
const phoneCallResponse = await client.call.registerPhoneCall({
  agent_id: 'oBeDLoLOeuAbiuaMFXRtDOLriTJ5tSxD',
}) ;
[(1) ]() В вашем случае
user_phone_number
сохраняется в metadata, чтобы позже использовать этот номер для обновления записи в Google Sheets или другой системе после завершения звонка. [(2) ](

▾

) Важно: metadata не используется для обработки во время звонка — это только для хранения. Для передачи данных агенту используйте
retell_llm_dynamic_variables
. [(1) ]() , [(4) ]()
​ To learn more, see the sources I considered: [[1] Register Phone Call](https://docs.retellai.com/api-references/register-phone-call) [[2] Personalization with Google Sheets](

https://www.youtube.com/watch?v=A9H3a33gM2M▾

) [[3] "No Metadata" but there is metadata](https://discord.com/channels/1186424376300679209/1412807904876302467) [[4] How can I create an outbound call](https://retell.discourse.group/t/how-can-i-create-an-outbound-call/132) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
d
в чем отличие от retell_llm_dynamic_variables
a
Thanks for the follow-up, @dropsofsteel. I'll get back to you soon.
### Answer ### Ответ Ключевые отличия: | |
metadata
|
retell_llm_dynamic_variables
| |---|---|---| | Назначение | Только для хранения | Инжектируется в промпт и описания инструментов | | Обработка | Не используется во время звонка | Используется агентом во время разговора | metadata — произвольный объект только для хранения. Можно сохранить внутренний ID клиента или другие данные. Не используется для обработки во время звонка. [(1) ]() retell_llm_dynamic_variables — динамические переменные в формате ключ-значение, которые инжектируются в промпт Response Engine и описания инструментов. , [(2) ]() Пример использования:
Copy code
javascript
retell_llm_dynamic_variables: {
  customer_name: "John Doe"
}
[(1) ]() Эта переменная
{{customer_name}}
будет доступна агенту во время разговора — например, для персонализированного приветствия. Когда что использовать: -
metadata
— для данных, которые нужны после звонка (ID клиента в CRM, номер телефона для обновления записи) -
retell_llm_dynamic_variables
— для данных, которые агент должен использовать во время разговора (имя клиента, сумма скидки и т.д.) [(3) ]()
​ To learn more, see the sources I considered: [[1] Register Phone Call](https://docs.retellai.com/api-references/register-phone-call) [[2] Get Call](https://docs.retellai.com/api-references/get-call) [[3] How can I create an outbound call](https://retell.discourse.group/t/how-can-i-create-an-outbound-call/132) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.