Ultralytics
Ultralytics powers YOLO26 vision AI models and the HUB platform for training and deployment. See 2026 pricing tiers and what YOLO26 adds.
About Ultralytics
What is Ultralytics (2026 update)
Ultralytics is a computer vision company behind the YOLO family of object detection models and the Ultralytics HUB platform, a cloud service for annotating data, training custom vision models, and deploying them to production or edge devices.
What is new in 2026
Ultralytics released YOLO26 in January 2026, its newest flagship model, built to be smaller, faster, and more accurate than earlier YOLO versions while running natively end-to-end without a separate non-maximum suppression (NMS) post-processing step. YOLO26 is built for edge and low-power devices and delivers up to 43% faster CPU inference thanks to a simpler architecture, technical changes include NMS-free end-to-end inference, DFL-free regression, a new MuSGD optimizer, Progressive Loss, and task-specific improvements for segmentation, pose estimation, and oriented bounding box detection. YOLO26 supports detection, instance segmentation, semantic segmentation, depth estimation, classification, pose estimation, and open-vocabulary versions of these tasks.
Ultralytics also held its YOLO Vision 2026 event on September 13 in Shenzhen and online, where it unveiled YOLO-VLM, the next flagship generation extending the YOLO family into vision-language tasks by pairing a lightweight YOLO front-end with a deeper language model layer for more efficient vision-language pipelines.
Key features
- YOLO26 model family for detection, segmentation, classification, pose estimation, and oriented bounding boxes
- NMS-free end-to-end inference for faster, simpler deployment
- Ultralytics HUB for dataset annotation, cloud training, and one-click deployment
- Manual, SAM3, and YOLO Smart annotation tools for labeling data
- Export to 20+ model formats for different deployment targets
- Access to 24 to 26 cloud GPU types depending on plan, including newer B200 and B300 GPUs
- YOLO-VLM, a new vision-language model family unveiled at YOLO Vision 2026
Pricing in 2026
Ultralytics HUB uses a free tier with usage limits plus a paid Pro tier billed per seat, and a custom Enterprise tier for organizations needing on-premise deployment or a commercial license to embed YOLO in closed-source products. Cloud GPU compute is billed separately by the hour depending on GPU type, starting at $0.24/hr with no markup or minimum commitment.
| Plan | Price | What you get |
|---|---|---|
| Free | $0/month | 100 GB storage, 100 models, 3 concurrent cloud training jobs, 3 cloud deployments, 24 cloud GPU types, $25 signup credit with a verified work email, AGPL 3.0 license |
| Pro | $29/seat/month | 500 GB storage, 500 models, 10 concurrent training jobs, 10 cloud deployments, team collaboration up to 5 members, 26 cloud GPU types including B200/B300, $30/seat/month in compute credits, priority support |
| Enterprise | Custom | Unlimited storage, models, deployments, and training jobs, SSO/SAML, on-premise deployment, ISO 27001 and SOC 2 compliance with SLA, Enterprise commercial license |
Who should use it
Ultralytics fits computer vision engineers and small ML teams who want a fast path from labeled data to a trained, deployed detection or segmentation model without building their own training pipeline. The Pro plan suits teams running frequent training jobs and deployments that exceed the free tier’s limits, while the Free plan works for individual developers experimenting with YOLO26 on smaller projects. Companies that need to embed YOLO models inside a closed-source commercial product need the Enterprise license rather than the default AGPL 3.0 license, which requires open-sourcing derivative code.
Limitations and alternatives
The default license on Free and Pro plans is AGPL 3.0, which has copyleft requirements that many commercial products cannot accept without buying an Enterprise license, so check the license terms before shipping a closed-source product built on YOLO. Cloud GPU costs are billed separately from the seat price and can add up quickly for heavy training workloads, check the official pricing page for current hourly GPU rates.
Real alternatives include Roboflow for a similar end-to-end computer vision platform with strong dataset management, Google Cloud Vertex AI for teams already on Google Cloud infrastructure, and Meta’s Detectron2 for open-source object detection without a hosted platform.
Frequently asked questions
What is YOLO26?
YOLO26 is Ultralytics's newest vision AI model, released in January 2026, offering NMS-free end-to-end inference, faster CPU performance, and support for detection, segmentation, classification, pose estimation, and oriented bounding boxes.
How much does Ultralytics HUB Pro cost?
The Pro plan costs $29 per seat per month and includes $30 per seat per month in cloud compute credits, 500 GB storage, and team collaboration for up to 5 members.
Do I need an Enterprise license to use YOLO commercially?
The Free and Pro plans use the AGPL 3.0 license, which has copyleft requirements. Organizations embedding YOLO in a closed-source commercial product need the separate Enterprise license.
What is YOLO-VLM?
YOLO-VLM is a new vision-language model family unveiled at Ultralytics's YOLO Vision 2026 event, combining a lightweight YOLO front-end with a deeper language model layer for vision-language tasks.
Information last verified: September 2026.
Related Tools
