yolo26n

by ultralytics

Edge-optimized NMS-free object detector with 43% faster CPU inference

8Kdl/month

108likes

~3M (Nano)params

HuggingFace Run on your data, free

Identifiers

Model ID

ultralytics/yolo26n

Feature URI

mixpeek://image_extractor@v1/ultralytics_yolo26n_v1

Overview

YOLO26 is Ultralytics' latest generation real-time object detector, engineered from the ground up for edge and low-power devices. It removes Decoupled Focal Loss (DFL) for simplified export, introduces end-to-end NMS-free inference for streamlined deployment, and uses ProgLoss + STAL for improved small-object accuracy. The MuSGD optimizer (SGD + Muon) delivers up to 43% faster CPU inference.

On Mixpeek, YOLO26 is the default object detection model for video analysis pipelines requiring real-time performance on edge hardware. Its NMS-free architecture eliminates a common deployment pain point, and the Nano variant runs on mobile and IoT devices while maintaining competitive detection accuracy.

Architecture

Attention-centric backbone with R-ELAN modules. NMS-free end-to-end inference via learned object queries. ProgLoss (progressive loss scaling) and STAL (Spatial-Temporal Attention Loss) for improved small-object detection. Available in Nano (N), Small (S), Medium (M), Large (L), and Extra Large (X) variants. Supports export to TensorRT, ONNX, CoreML, TFLite, and OpenVINO.

Mixpeek SDK Integration

import { Mixpeek } from "mixpeek";

const mx = new Mixpeek({ apiKey: "API_KEY" });

// Managed: create a collection over a bucket; Mixpeek runs this model's extractor
const collection = await mx.collections.create({
  namespace_id: "my-namespace",
  collection_name: "my-collection",
  source: { type: "bucket", bucket_ids: ["bkt_your_bucket"] },
  feature_extractor: {
    feature_extractor_name: "s3",
    version: "v1",
    parameters: { model_id: "mixpeek://image_extractor@v1/ultralytics_yolo26n_v1" },
  },
});