Model library

Every top model, one canvas

The best model changes every few months. Treza puts them all in one picker, on one credit balance, behind one API, so switching is a dropdown, not a migration.

Also on the canvas

Language, speech, and embedding models power the prompting, narration, and search stages of a pipeline. And any Hugging Face or OpenRouter model id can be typed straight into a node.

Language

Drafting, rewriting, classification, and reasoning with open-weight models.

  • MetaLlama 3.3 70B
  • DeepSeekDeepSeek R1
  • QwenQwen2.5 72B
  • MistralMixtral 8x22B

Speech & audio

Voiceovers, narration, transcription, and embeddings for search and retrieval.

  • HEKokoro 82M
  • MistralVoxtral Mini TTS
  • GeminiGemini Flash TTS
  • OpenAIWhisper Large v3
  • BGBGE Large EN