Custom AI Models from a Prompt, Now with One-Click Managed Inference

Oumi now lets teams move from a prompt to a hosted custom model in one workflow.

It has been just over a month since the Oumi Platform launched. More than 2,000 users have signed up and started applying custom models to real business problems. These are the latest platform updates.

One-click managed inference

Teams can host fine-tuned models directly on the Oumi Platform and start inference immediately. Open the Deployments tab, select Create Deployment, and follow the setup flow.

Each deployment exposes an OpenAI-compatible endpoint that teams can access with an Oumi API key. On-demand deployment is billed per second, with A100 80GB instances at $3 per hour and H100 or H200 instances at $7 per hour.

New open models

The platform now supports Qwen 3.6 and 3.5 27B with LoRA fine-tuning, plus Gemma 4 2B and 4B with full-weight fine-tuning. If a model is missing, the team can help evaluate the right addition.

Better data quality checks

Oumi now flags conversations with empty or poorly formatted messages. That makes it easier to find noisy or invalid examples before using a dataset for evaluation, synthesis, or training.