Dyad0.21.0
More models - turbo models, free models, and much more!
Dyad v0.21.0 brings support for many more models! Dyad Pro members get access to Turbo models which are 4-10x faster than regular models! There are also 2 new free models from OpenRouter, support for 3 more AI providers, and much more!
Highlights
Turbo models (Pro)
This video is uncut and not sped up! Turbo models are 4-10x faster at generating responses than regular models.
If you're tired of waiting for AI responses to slowly stream in, I think you'll like Turbo models. I've been using these a lot personally because I'm impatient :) Under the hood, Turbo models use top open-source models (Qwen3 Coder and Kimi K2) served by the fastest AI providers: Cerebras and Groq. We've also sped up Smart Context when using Turbo models, so everything feels fast. Get access with Dyad Pro today!
More free models
We added two more free models from OpenRouter: Qwen3 Coder and DeepSeek v3.1, and added a new model option called Free (OpenRouter) which will cycle through all the free OpenRouter models and let you use whichever one is available.
As always, we're committed to making Dyad the best free AI app builder!
AWS Bedrock, GCP Vertex, and xAI provider support
A big shoutout to the Dyad community for adding 3 new AI providers:
- Google Vertex AI provider by @rakibulrocky14 in #1163
- Amazon Bedrock provider by @samratjha96 in #1185
- xAI (Grok) as AI provider by @princeaden1 in #1209
Chat search
If you've been trying to find an old chat for an app with dozens of chats, you're going to appreciate the new chat search feature, which lets you search through all the chats and messages for your current Dyad app. Thanks, @iamevansobeng, for working on this feature!
New community Expo template
Thanks, @shaaraa, for adding a new community Expo template in #1177!
Expo lets you build mobile and web apps using React Native! If you've been having trouble setting up Capacitor, you can check out Expo, which has a simpler development setup.
Other improvements
There are also a few smaller updates to Dyad:
- Claude Sonnet 4 now supports 1M tokens! (Keep in mind you'll be charged more for >200k tokens)
- PHP files are now included in the AI context window
Feedback
If you run into any issues or have suggestions for this release, the best way to share them is by reporting a bug and uploading a chat session.