Oumi
Build, fine-tune, evaluate, and deploy specialized foundation models with an open-source stack and managed platform.
About Oumi
Oumi is an end-to-end platform for ML engineers, researchers, and teams that need to build specialized foundation models without stitching together separate tools for data preparation, training, evaluation, and deployment. Its open-source stack uses a unified, configuration-driven interface that can move from a single local GPU to a cluster, while the hosted platform adds agent-led automation, managed training, evaluation, synthesis, and autoscaling inference. The Free plan gives new users up to $50 in hosted credits for three months. Pro costs $25 per month, includes the same amount in monthly platform credits, and switches to pay-as-you-go billing after those credits. Enterprise adds private deployment options, dedicated capacity, SLAs, and hands-on support. Usage charges also apply to hosted tools, storage, training, and inference, so teams should estimate token, GPU-hour, and uptime costs before moving a workload into production.
Best for
Best use cases
Fine-tune a language model on one GPU or a cluster
Synthesize and curate training data before evaluation
Deploy a fine-tuned or LoRA model for production inference
Compare before choosing
Choose Free to test the hosted workflow with temporary credits
Choose Pro for concurrent jobs, distillation, and autoscaling inference
Choose Enterprise for private infrastructure, capacity guarantees, and SLAs
Features
Unified model workflow
Prepare data, train or fine-tune foundation models, evaluate results, and deploy through one configuration-driven stack.
Open-source stack
Run Oumi OSS locally or across clusters with an open-source codebase, more than 200 recipes, and support for more than 100 models.
Hosted automation
Use the hosted Oumi agent for data synthesis, analysis, curation, evaluation, SFT, PEFT, and production inference.
Pros
- A unified configuration can move from a local GPU to clustered training.
- Open-source tooling keeps self-managed development available.
- The free hosted tier includes starter credits for testing.
Cons
- Metered hosted services can make production costs workload-dependent.
- Pro limits model-weight downloads to one per month.
- Private deployment and SLAs sit behind custom Enterprise pricing.
Pricing Plans
Free
- Up to $50 in credits for 3 months
- Oumi agent automation
- Data synthesis, analysis, and curation
- Open- and closed-model evaluation
- SFT and PEFT training
Pro
- Everything in Free
- On-policy distillation
- Multiple concurrent jobs
- One model-weight download per month
- Production inference with autoscaling
- $25 monthly platform credits
Enterprise
- Everything in Pro
- BYOC, VPC, and on-premises deployment
- Dedicated capacity scaling to thousands of GPUs
- Guaranteed throughput and SLAs
- Advanced training methods and custom pipelines
- Dedicated experts
* Pricing information is for reference only. Please refer to the latest announcement on Oumi's official website.