Skip to main content

Overview

Moondream is an open-weight family of Vision Language Models (VLMs) built for powerful, efficient visual reasoning. The Moondream 3 family uses a mixture-of-experts architecture with grounded visual reasoning, a 32k context window, and native support for multiple vision skills—like pointing, counting, and object detection—all designed with a deployment-friendly ethos.

For local inference, Photon exposes Moondream, Qwen, Gemma, and speech-to-text models including Whisper, Qwen3-ASR, and Parakeet through the Moondream Python package. For hosted inference, choose a model through the Cloud API quickstart.

Key stats​

  • 9B total params, 2B active params (maintains similar inference speeds to our previous models)
  • 32k context window (up from 2k)

Model Skills​

Moondream has built-in vision-specific skills that make it easy to generate specific types of vision outputs (e.g., bounding boxes, or 2D points). These are:

  • Object Detection
  • Pointing and Counting
  • Visual Question Answering
  • Captioning
  • Image Segmentation

Performance Benchmarks​

Selected benchmark results are shown alongside larger frontier models for comparison. Moondream is designed to produce strong results with substantially lower inference cost.

TaskMoondream 3 PreviewGPT 5Gemini 2.5-FlashClaude 4 Sonnet
Object Detection
RefCOCOg88.649.875.126.2
RefCOCO+81.846.370.223.4
RefCOCO91.157.275.830.1
Counting
CountBenchQA93.289.381.290.1
Document Understanding
ChartQA86.685*79.574.3*
DocVQA88.389*94.289.5*
Hallucination (higher is better)
POPE89.088.488.184.6

License​

Copyright (c) 2025 M87 Labs, Inc. This distribution includes Model Weights licensed under the Business Source License 1.1 with an Additional Use Grant (No Third-Party Service). Commercial hosting or rehosting requires an agreement with contact@m87.ai.


Learn more​