Free web
Free source (opens in a new tab)
Month
For: Building a first model
Limits
- Up to $50 in credits for 3 months
- No credit card required
Includes
- Oumi Agent automation
- Data synthesis
2 more
- Evaluation with failure modes
- SFT and LoRA/QLoRA
CodingDocumented overviewLast checked Oct 1, 2026
Oumi: Pricing, Features & Best Use Cases
Platform and agent for building, evaluating, fine-tuning and deploying custom AI models from a natural-language prompt.
(open full size in a new tab)
(open full size in a new tab)
(open full size in a new tab)
(open full size in a new tab)
(open full size in a new tab)
Our take: Choose Oumi if you want an agent to take a task prompt through data synthesis, evaluation, fine-tuning and deployment of a small custom model that you own.
Main limitation: Free credits expire after 3 months and Pro limits model weight downloads to one per month.
Who Oumi fits
Choose Oumi if you want an agent to take a task prompt through data synthesis, evaluation, fine-tuning and deployment of a small custom model that you own. Based on the vendor's documentation, not on a hands-on test.
Best for: ML teams building custom models for specific tasks
Also a good fit for
- Enterprises wanting to own rather than rent AI models
- Developers experimenting with fine-tuning and distillation
Consider another tool if free credits expire after 3 months
Or if
- Pro plan is $25/month plus pay-as-you-go usage
- Pro allows only 1 model weight download per month
Oumi in the news
Headlines from US outlets, selected automatically — not endorsements. Updated Oct 2, 2026
Oumi pricing and usage limits
Prices checked Oct 1, 2026 on the official US pricing page. Official pricing ↗ Prices and limits change often.
Free access and trial conditions
Up to $50 credits for 3 months (work email vs $25 personal per research notes).
Paid plans: what changes as you upgrade
Pro web
$25/mo source (opens in a new tab)
Month
For: Professional model development
Limits
- $25/month platform credits
- 1 weight download per month
Includes
- On-policy distillation
- Concurrent jobs
1 more
- Production inference with autoscaling
Enterprise web
Custom pricing source (opens in a new tab)
Unknown
For: Production at scale
Limits
- Contract terms
Includes
- BYOC VPC and on-prem
- Dedicated GPU capacity
2 more
- SLAs
- RL and custom pipelines
Usage caps, credits and extra costs
Usage: agent tokens $6.25 in / $31.25 out per 1M; SFT from $0.49 per 1M tokens; H100 $4.00/GPU-hr.
What is Oumi, and how does it work?
Oumi is a hosted platform for developing custom AI models. Starting from a prompt, the Oumi Agent automates evaluation, test-set creation, data synthesis, training method selection (SFT, LoRA, QLoRA, on-policy distillation) and one-click deployment to GPUs that scale to zero when idle, then retrains on production feedback. The platform targets organizations that want to own task-specific models instead of renting frontier models, with example use cases in banking, insurance, healthcare, retail and customer support. Documentation covers datasets, evaluators, training, deployment and a model playground, and users can bring their own OpenAI, Anthropic or Google API keys.
What you provide and what you get back
You provide
- Natural-language task prompt
- Documents such as PDFs
- Datasets and evaluators
You get back
- Fine-tuned model
- Evaluation reports with failure modes
- Synthetic training rows
- Hosted inference endpoint
Where you can use it
- Platforms
- Web
- Account
- Required
- Training
- SFT, LoRA, QLoRA, on-policy distillation
- GPUs
- A100 $2.90/hr, H100 $4.00/hr
- Agent tokens
- $6.25 in / $31.25 out per 1M
- Storage
- 4 GB per month per $1
What people use Oumi for
Scenarios are worked out by our editors from the documentation; the actual result depends on your input and plan.
Distill a compact task-specific model from a stronger model
- Who
- ML and product teams
- You provide
- A stronger model plus task prompt
- Expected result
- A compact distilled model for that task
- Plan needed
- Pro for on-policy distillation
- Limitation
- Training runs on 8 GPUs, subject to availability source (opens in a new tab)
Evaluate a model on a custom dataset and surface failure modes
- Who
- Teams checking quality
- You provide
- Custom dataset
- Expected result
- Failure-mode analysis of the model
- Plan needed
- Free
- Limitation
- Judgments billed at 1,000 per $1 source (opens in a new tab)
Synthesize training data from documents such as PDFs
- Who
- Domain teams with documents
- You provide
- PDFs
- Expected result
- Synthetic training data rows
- Plan needed
- Free
- Limitation
- Synthesis billed at 1,000 rows per $1 source (opens in a new tab)
Key features and their practical limits
Oumi Agent that runs ML workflows from natural-language prompts
Runs evaluation, data, training and deployment steps from a natural-language prompt. Cuts manual ML work.
Automated evaluation with failure-mode analysis
Scores models on your dataset and clusters failures. Shows what to fix.
Data synthesis and curation with open or closed models
Generates and curates rows with open or closed models. Builds training sets without labeling.
SFT, LoRA/QLoRA and on-policy distillation training
Trains with SFT, LoRA, QLoRA and on-policy distillation. Chooses suitable training methods.
Oumi pros and cons
Where Oumi is strong
- Free tier with credits and no credit card required You can start without payment details. documented source (opens in a new tab)
- Transparent per-token, per-GPU-hour and per-job usage pricing in USD Cost can be estimated per job. documented source (opens in a new tab)
- Detailed public documentation with quickstart and CLI reference Developers can follow guides and CLI reference. documented source (opens in a new tab)
- Enterprise option with VPC and on-prem deployment Regulated teams can keep data in their cloud. documented source (opens in a new tab)
Limitations to check before paying
- Free credits expire after 3 months Unused free credit disappears. documented source (opens in a new tab)
- Pro plan is $25/month plus pay-as-you-go usage Heavy users pay on top of the $25. documented source (opens in a new tab)
- Pro allows only 1 model weight download per month Owning weights is restricted on Pro. documented source (opens in a new tab)
- Production inference billed per GPU-hour for uptime Idle uptime costs money. documented source (opens in a new tab)
How to use Oumi for fine-tuning a task-specific model
Set up Oumi
- Sign up free at the Oumi platform.
- Describe the task in a prompt to the Oumi Agent.
- Upload documents or a dataset.
- Choose a base model.
Example: fine-tuning a task-specific model
- Have the agent synthesize training rows from your PDFs.
- Run an evaluation to find failure modes.
- Start an SFT or LoRA job.
- Deploy to an inference endpoint.
Review and export the result
- Re-run the evaluation on the trained model.
- Review credit usage.
Fine-tune a small model to classify insurance claim emails into five categories using my labeled PDFs.
Oumi integrations and API
Native integrations and export formats
Export: Model weights (1/month on Pro)
API access and separate billing
Oumi has no public API in its official documentation. API access not confirmed in loaded pages. API documentation ↗
Privacy, data use and commercial rights
Summarized from the vendor's current policies; this is not legal advice and a vendor's badge is not an independent audit.
How your inputs and outputs are handled
Datasets and prompts are processed on the platform.
Commercial use and plan restrictions
VPC and on-prem only on Enterprise.
Controls for business users
- BYOC VPC and on-prem
Oumi alternatives by need
"Better suited" is based on documented features and pricing, not on a head-to-head test.
Replit: better suited to
GitHub Copilot: better suited to
Cursor: better suited to
Explore Oumi further
Related to Oumi
More coding tools
Guides and comparisons
Oumi user reviews
No user reviews yet.
Used Oumi? Be the first to share your experience.
Write a review
Oumi FAQ
How long do free credits last?
Up to $50 in credits for 3 months, no credit card required. source (opens in a new tab)
What does Pro include?
$25 per month with $25 of platform credits, on-policy distillation, concurrent jobs, production inference and one weight download a month. source (opens in a new tab)
What does supervised fine-tuning cost?
From $0.49 per 1M training tokens for models up to 16B, rising to $6.00 for 80-300B. source (opens in a new tab)
Do I pay for OpenAI or Anthropic inference?
Not through Oumi if you add your own key; you pay the provider directly. source (opens in a new tab)
Can I deploy in my own cloud?
Enterprise offers BYOC, VPC and on-prem. source (opens in a new tab)
Sources and review methodology
Official sources and verification dates
- Oumi – Your AI factory homepage · checked Oct 1, 2026
- Pricing – Oumi homepage · checked Oct 1, 2026
- WELCOME TO OUMI - Oumi Docs homepage · checked Oct 1, 2026
Not verified: data retention; training use of data; API availability.
How we documented and maintain this page
Every fact about Oumi above comes from the 3 official pages listed here, last compared on Oct 1, 2026; we have not run Oumi ourselves, so it has no editorial score. How we check and re-check pages.
Article updated Oct 10, 2026; prices checked Oct 1, 2026. Outbound links to vendors are plain links; we do not earn commissions. How we review · Report an error