YOLOE
Real-time 'see anything' detection and segmentation with text, visual, or prompt-free modes in one model.
About
YOLOE (Real-Time Seeing Anything) comes from Tsinghua's MIG group and pushes open-world YOLO past YOLO-World. A single model handles three prompt styles: text prompts through a re-parameterizable region-text alignment module, visual prompts via a semantic-activated visual prompt encoder, and a prompt-free mode that uses a built-in large vocabulary so you can run it with no language input at all. It does detection and segmentation together, and the re-parameterization tricks keep inference as fast as a plain YOLO.
On LVIS, the authors report YOLOE-v8-S beating YOLO-Worldv2-S by 3.5 AP with 3x less training cost and 1.4x faster inference, and YOLOE-v8-L transfers to COCO slightly above closed-set YOLOv8-L with roughly 4x less training time. It was accepted to ICCV 2025. The research code is at THU-MIG/yoloe and Ultralytics folds YOLOE into its framework, which carries AGPL-3.0 obligations, so the practical licensing story needs a careful read before commercial use.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Author-reported (arXiv:2503.07465): YOLOE-v8-S +3.5 AP over YOLO-Worldv2-S on LVIS at 3x less training cost and 1.4x inference speedup
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: arXiv paper (2503.07465, ICCV 2025) with LVIS numbers from the abstract; research repo confirmed at github.com/THU-MIG/yoloe. Strengths: unifies text, visual, and prompt-free modes; strong reported efficiency gains over YOLO-Worldv2. Docked for: (1) repo license not explicitly confirmed in the sources checked, and the Ultralytics distribution is AGPL-3.0, creating commercial-use uncertainty; (2) newer model with less independent evaluation than YOLO-World. License: not explicitly confirmed — flag for review.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 7.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · ICCV 2025 (arXiv 2503.07465, Mar 2025)
- · YOLOE-v8-S +3.5 AP over YOLO-Worldv2-S on LVIS at 3x less training cost, author-reported
- · Unifies text, visual, and prompt-free prompt modes