Not runnable todayDetectionVerified 2026-11-05
YOLO-World v2 (Ultralytics)
yolov8s-worldv2.pt
YOLO-World v2 (Ultralytics) is a image and text object detection model (AGPL-3.0). Not runnable on Mixpeek today.
Something blocks it today, such as the license, remote code or a GPU-only runtime. The reason is below.
No demo for YOLO-World v2 (Ultralytics) yet
A video, image, slideshow or playground goes here once we record one on our own data.
What it is good and bad at
- Best at
- Open-vocabulary detection from a list of text class names, in real time; set classes per namespace without training.
- Worst at
- Rare or compositional phrases; accuracy below closed-set YOLO on COCO classes; license blocks closed commercial use.
- Task
- Detection
- Input
- Image, Text
- Output
- Boxes, Labels, Scores
- Use cases
- Object detection, Brand detection
Templates that use it
No template uses this model yet. Browse templates.
Input and output
Input
- image
- resolution
- 640x640 letterboxed
- prompts
- list of class names (CLIP text encoder), set once with set_classes
Output
- boxes
- xyxy pixels
- labels
- open vocabulary (the given class list)
- scores
- 0..1
How Mixpeek wires it
- Extractors
- None today
- Feature URIs
- None today
- Vector indexes
- None
- Retriever stages
- None
- Combines with
- taxonomy labels -> YOLO-World class list -> boxes -> crops -> SigLIP
- detector boxes -> crops -> SigLIP vectors per object (one document per object, or a multi-vector per frame)
How it runs
- Status
- Not runnable today Blocked by AGPL-3.0 and pickle .pt weights. Would run as a custom extractor plugin that exports the checkpoint to ONNX once (ultralytics export format=onnx) and serves it with onnxruntime on CPU, emitting boxes, labels and crops. The class list is baked in at export.
- Serve apps
- None
- Runtime
- ultralytics YOLOWorld
- Weights
- Pickle, no remote code
- Parameters
- Not recorded
- Size on disk
- Not recorded
- Preprocessing
- Letterbox to 640x640 (stride-aligned padding), RGB, scale to 0..1.
- Postprocessing
- Confidence threshold then NMS.
- Serving
- both
- Hardware
- Runs on CPU (slower without a GPU)
Scale to zero
- Minimum replicas
- 0
- Maximum replicas
- 1
- Idle timeout
- 600 s
- Memory
- 2 GB
- CPUs
- 1
- GPUs
- 0
- Cold start, node warm
- Not recorded
- Cold start, new node
- Not recorded
Measured cost
Unmeasured
No committed measurement; plan row 3 measures it in the standalone image.
Our evaluations
Unmeasured
We have not evaluated this model on our own data yet. Publisher results are in the sources below.
Questions
- What is YOLO-World v2 (Ultralytics) best at?
- Open-vocabulary detection from a list of text class names, in real time; set classes per namespace without training.
- What are the limits of YOLO-World v2 (Ultralytics)?
- Rare or compositional phrases; accuracy below closed-set YOLO on COCO classes; license blocks closed commercial use.
- Can YOLO-World v2 (Ultralytics) be used commercially?
- No. AGPL-3.0: offering it over a network obliges publishing the whole application's source under AGPL; closed commercial use needs an Ultralytics Enterprise license. The original Tencent AILab repo is GPL-3.0.
- How does YOLO-World v2 (Ultralytics) run on Mixpeek?
- Blocked by AGPL-3.0 and pickle .pt weights. Would run as a custom extractor plugin that exports the checkpoint to ONNX once (ultralytics export format=onnx) and serves it with onnxruntime on CPU, emitting boxes, labels and crops. The class list is baked in at export.