Lovable: GPT-6 Astra, Gemini 3.8 Flash, and Gemini 3.5 Transcribe Added
LovableLovable added three new models that builders can wire into their apps' own AI features: GPT-6 Astra, OpenAI's premium-priced model for complex reasoning, coding, research, and document creation; Gemini 3.8 Flash, now the default chat model, which accepts text, image, audio, and video input; and Gemini 3.5 Transcribe, the new default speech-to-text model that auto-detects spoken language across more than 85 languages. Builders can either explicitly request one of these models or describe what they need and let Lovable choose.
Key Takeaways
- Lovable added GPT-6 Astra for apps that need premium, high-capability reasoning, coding, research, or document creation.
- Gemini 3.8 Flash is now the default chat model for app-level AI features and supports text, image, audio, and video input.
- Gemini 3.5 Transcribe becomes the default speech-to-text model, with automatic language detection across 85+ languages.
- These models power AI features inside apps built on Lovable, distinct from the model Lovable itself uses to generate code.
- Builders can either name the specific model they want or simply describe the desired behavior and let Lovable choose.
- The update keeps Lovable's app-facing model lineup current with the latest releases from both OpenAI and Google.
Sources & Mentions
1 external resource covering this update
Three New Models for App-Level AI Features
Lovable expanded the roster of models available for the AI features built into apps made on the platform, not the model Lovable itself uses to build, but the ones an app's own chatbots, assistants, or transcription tools can call at runtime. The update adds three options spanning three different jobs: heavyweight reasoning, everyday multimodal chat, and speech-to-text.
GPT-6 Astra for Hard Problems
GPT-6 Astra is positioned as OpenAI's most capable model for demanding work: complex reasoning, coding, research, and document creation. It comes at a premium price, making it the option to reach for when a task genuinely needs the strongest available model rather than the fastest or cheapest one.
Gemini 3.8 Flash Becomes the Default Chat Model
Gemini 3.8 Flash takes over as Lovable's new default chat model for app-level AI features. It accepts text, image, audio, and video input, giving builders a multimodal default without needing to configure anything extra for apps that just need solid, general-purpose conversational AI.
Gemini 3.5 Transcribe for Speech-to-Text
Gemini 3.5 Transcribe becomes the new default speech-to-text model, with automatic spoken-language detection across more than 85 languages. This is aimed at apps that need to turn voice input into text, from voice notes to multilingual customer interactions, without builders having to manually specify a language.
Choosing a Model
Builders do not need to know these model names to benefit from them: asking Lovable to use a specific one works, but so does simply describing the desired behavior and letting Lovable pick the appropriate model automatically. This keeps the barrier to adopting newer, more capable models low for builders who would rather describe outcomes than manage model selection themselves.