External Models
Mixtrain integrates with leading AI model providers for image generation, video generation, and multimodal tasks. Add a provider once, and all its models become available across your workflows, evaluations, and routers.
Supported Providers
Image & Video Generation
| Provider | Models | Description |
|---|---|---|
| Runway | Gen-4.5, GWM-1 | Video generation and world models |
| Luma Labs | Ray3.14, Ray3 Modify | Video, 3D, and HDR generation |
| Black Forest Labs | FLUX.2 Pro, FLUX.2 [klein] | High-quality image generation and editing |
| fal | FLUX.2, Kling 2.6, LTX 2.0 | 1000+ image, video, and audio models |
| Veo 3.1, Genie | Video generation and interactive 3D worlds |
Language & Multimodal
| Provider | Models | Description |
|---|---|---|
| Gemini 3.1 Pro, Gemini 3 Deep Think | Multimodal reasoning and agentic AI | |
| Anthropic | Claude Opus 4.6, Sonnet 4.6 | Advanced reasoning and coding |
| OpenAI | GPT-5.4, GPT-5 mini | Language and image models |
Quick Start
Add a provider and start generating videos in a few lines:
mixtrain provider add runwayfrom mixtrain import Model
model = Model("gen-4.5")
result = model.run({
"prompt": "A drone shot flying over a mountain lake at sunrise",
"duration": 4
})How It Works
- Add a provider with your API key using the CLI or SDK.
- Browse available models with
mixtrain model catalog. - Use models anywhere in Mixtrain — workflows, evaluations, routers, or direct calls.
External models work seamlessly alongside your own fine-tuned models, so you can mix and match providers in a single workflow or router.
See Also
- Robotics Models - Fine-tune GR00T N1.6, pi0.5, SmolVLA
- Vision Models - Fine-tune SmolVLM 2, SigLIP 2, SAM 2.1
- Audio Models - Fine-tune Parakeet, Canary Qwen, Whisper