> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.vapi.ai/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.vapi.ai/_mcp/server.

# What's New: Week of October 19, 2026

## New Models and Voices

1. **Claude Haiku 5.5**: The [`claude-haiku-5-5`](/providers/model/anthropic) model is now available for assistants and squads, with a reasoning-effort control, and it can be selected as an eval judge model.
2. **Gemini 3.8 Flash**: The [`gemini-3.8-flash`](/providers/model/gemini) model is now available on the Google provider.
3. **Cartesia Sonic-3.6**: The Cartesia [`sonic-3.6`](/providers/voice/cartesia) voice is now available in the dashboard voice settings.

## Features

1. **Priority processing (OpenAI Fast mode)**: A new [Priority processing](/providers/model/openai#service-tier) switch in the assistant editor runs supported OpenAI models in OpenAI's Fast service tier, billed at OpenAI's Fast rates, and the served service tier now appears in `call.costs`.
2. **Monitor failed server messages**: You can now add a "Server message failed" monitor that alerts you when a server message (webhook) fails to deliver, filterable by message type, URL, status code, and failure reason.

## Fixes and Improvements

1. **Twilio SIP warm-transfer fix**: On Twilio SIP calls, the warm-transfer URL is now set on the `<Sip>` verb instead of `<Dial>`.
2. **GPT-Live silence fix**: GPT-Live now drops queued silence so voice simulations no longer overflow.
3. **Call recording reliability**: Recording update and token creation on the Daily media provider are now retried, so video recordings are more reliable.