Whats new for 9-21-26

fern/changelog/2026-09-21.mdx

@@ -0,0 +1,5 @@
+# What's New: Week of September 21, 2026
+
+1. **Faster OpenAI Responses with Service Tiers**: You can now set `serviceTier` to `fast` on your OpenAI [GPT-5.6 and GPT-5.5 models](/providers/model/openai#service-tier) to request OpenAI's faster processing tier for latency-sensitive calls.
+
+2. **Provider Request IDs for Bring Your Own Key (BYOK)**: For [BYOK](/customization/provider-keys) models, call logs now include the provider's request ID for your language model, transcriber, and voice providers, so you can share it with the provider when getting support.

fern/providers/model/openai.mdx

@@ -51,6 +51,27 @@ For additional API configuration options, review the [`OpenAIModel` fields](/api
 | GPT Realtime 2 | `gpt-realtime-2` |
 | o3 | `o3` |
 
+## Service tier
+
+For latency-sensitive assistants, you can request OpenAI's faster processing tier by setting `model.serviceTier` on a supported model.
+
+| Value | Behavior |
+| --- | --- |
+| `fast` | Requests OpenAI's fast processing tier for lower latency, billed at OpenAI's fast-tier rates. If OpenAI downgrades the request under load, the call is billed at standard rates. |
+| `auto` | Uses the service tier configured for your OpenAI project. |
+| `default` | Standard processing and billing. |
+
+`serviceTier` applies only to models that support fast processing: `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna`, and `gpt-5.5`. It is ignored for other models. When unset, Vapi uses the service tier configured for your OpenAI project. See [OpenAI's pricing](https://openai.com/api/pricing) for fast-tier rates.
+
+```bash
+curl -X PATCH "https://api.vapi.ai/assistant/ASSISTANT_ID" \
+  -H "Authorization: Bearer $VAPI_API_KEY" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": { "provider": "openai", "model": "gpt-5.6-sol", "serviceTier": "fast" }
+  }'
+```
+
 ## Related
 
 <CardGroup cols={2}>