Oumi
An open-source LLM training and deployment platform under Apache 2.0. Build specialized models on your own infrastructure or use its hosted service.
Oumi builds specialized AI models for teams that want control over their training data, model weights, and deployment. Its Apache 2.0 open-source stack runs on laptops, clusters, and your own servers, while its hosted service automates model development from a plain-English task description. You own the resulting weights, data, and training recipes.
The platform covers data preparation, fine-tuning, evaluation, and inference in one workflow. It supports text and vision-language models, including Qwen, Llama, Gemma, DeepSeek, and gpt-oss. ML researchers can customize individual components and reproduce experiments; teams building production models can use existing recipes rather than write their own training loops.
Oumi can generate training examples, curate data with LLM judges, and compare models against benchmarks or custom evaluations. Training methods include LoRA, QLoRA, and reinforcement learning, with distributed training for larger workloads. For serving models, it works with vLLM and SGLang. It also connects to commercial APIs such as OpenAI, Anthropic, and Vertex AI.
The hosted service automates model selection and training choices, then uses production feedback to find failures and retrain on a schedule you control. Evaluation results, training examples, and recipes remain inspectable and editable. Deployment can use Oumi's cloud GPUs or your own infrastructure, including on-premises servers and a private cloud. Specialized models can also run on smartphones and embedded devices.
