New Model
Just released yesterday, but we are already impressed with its performance.
Starting today, GPT-4o mini will become the default model at Chatwith
, superseding GPT-3.5 Turbo. It's almost as fast and capable as GPT-4o but at much lower cost, making it the best replacement to GPT-3.5.This model is available to all our customers on all plans. We recommend upgrading as soon as possible to benefit from increased capabilities. An automatic upgrade of all chatbots on GPT-3.5 will happen in the following weeks.
We recommend testing your chatbot after the change to make sure it works as expected on the new model. Please contact our support if you experience any issues.
Small changes and fixes
Here's a list of other things that work better now:
- Fix: Missing dollar ($) signs in some messages
- Fix: Appearance section not possible to save with default welcome message
- Fix: Missing "required" information from OpenAPI specification in Actions
- Fix: Knowledge source character count not refreshing after deletion
- Fix: Source training status not always visible in onboarding
- Fix: Dialog height preventing more payload fields in Webhook Actions
- Fix: Email reports not sending for some customers
- Improve: Don't show close icon when chatbot is embedded
- Improve: Add current day of week to chatbot context
- Improve: Scroll sources table in onboarding
- Improve: Better indication a custom domain needs extra verification
- Improve: Allow testing webhooks that don't return JSON
- Improve: Add Botfather link in the Telegram form