DeepFellow DOCS

Supported Models

DeepFellow as a framework doesn't have any limitation as to which models you can use, so pick from thousands of models of various types – LLM, STT, TTS, txt2img, embedding, etc.

You can also use proprietary models or connect your own, to build exactly what you need.

Install any models you want through supported backends:


Below there's a list of the most popular specialized open-source models available to use with DeepFellow.

Community offers hundreds of thousands more models that you can also use.

Text, Reasoning, Tool Use

  • gpt-oss
  • deepseek-r1
  • qwen3
  • mistral
  • gemma3
  • llama3.1
  • phi4
  • polish models: bielik, pllum
  • and many more

Embedding

  • embeddinggemma
  • mxbai-embed-large
  • nomic-embed-text
  • and many more

Image Generation

For image generation, DeepFellow supports stable diffusion models from HuggingFace and CivitAI.

They can be layered with LORAs (low-rank adapters) to adjust the artistic style of the image being generated.

Read image generation guide to learn more.

Speech-to-Text/Text-to-Speech

  • faster-whisper
  • piper
  • Kokoro
  • XTTS-v2
  • Chatterbox TTS
  • VibeVoice
  • pyannote/speaker-diarization-3.1
  • nvidia/parakeet-tdt-0.6b-v2
  • and many more

Read our guides to learn more about Speech-to-Text / Text-to-Speech:

Adding More Models

DeepFellow allows you to install your own custom models or connect any application packaged as a Docker container.

You can install:

  • Models available in the Ollama library – use ollama service in DeepFellow Infra.
  • Models available in HuggingFace in GGUF format – use llamacpp service in DeepFellow Infra.
  • Models available in HuggingFace supported by vLLM – use vllm service in DeepFellow Infra.
  • Models from OpenAI/Google – use openai/google service in DeepFellow Infra.
  • Any image generation model compatible with stable diffusion (e.g. Civitai, HuggingFace) – use stable-diffusion service in DeepFellow Infra.
  • Any LoRA compatible with stable diffusion (e.g. HuggingFace, Civitai) – use stable-diffusion service in DeepFellow Infra.
  • Any docker image – use custom service in DeepFellow Infra.

Learn more about Custom Models.

Proprietary Models

We offer an anonymization layer for using commercial models from OpenAI and Google:

We use cookies on our website. We use them to ensure proper functioning of the site and, if you agree, for purposes such as analytics, marketing, and targeting ads.