This model is a fine-tuned version of Qwen/Qwen3.5-9B-Base on the Omni-Edu-70K dataset. The following hyperparameters were used during training: - learningrate: 5e-06 - trainbatchsize: 1 - evalbatchsize: 8 - distributedtype: multi-GPU - numdevices: 8 - gradientaccumulationsteps: 8 - totaltrainbatchsize: 64 - totalevalbatchsize: 64 - lrschedulertype: cosine - lrschedulerwarmupsteps: 0.1 - numepochs: 3.0 - Transformers 5.2.0 - Pytorch 2.10.0 - Datasets 4.0.0 - Tokenizers 0.22.2
Open-weight model · Image and text to text
moondream3.1-9B-A2B
by Moondream moondream/moondream3.1-9B-A2B
Moondream 3.1 is a vision language model with a mixture-of-experts architecture (9B total parameters, 2B active). It delivers state-of-the-art visual reasoning and detection while staying fast and cheap to deploy.
Parameters9.3B
Context—
Weights10.5 GB
Licenseother
AccessOpen weights
Monthly Downloads12.1k
Runs On
What it takes to serve moondream3.1-9B-A2B (9.3B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 18.5 GB | 22.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 9.3 GB | 11.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 4.6 GB | 5.6 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.
Model Card
Moondream 3.1 is a vision language model with a mixture-of-experts architecture (9B total parameters, 2B active). It delivers state-of-the-art visual reasoning and detection while staying fast and cheap to deploy. Skills include query, detect, point, and caption, all native and all returning structured output. For the full story on what's new — including how we trained it and how it holds up on your own tasks — see the release notes. Photon is Moondream's high-performance inference engine. It runs the model locally on NVIDIA GPUs (Ampere or newer) and Apple Silicon Macs, with the same API as Moondream Cloud. No API key is required to run the base model locally. (You'll only need one to run…
Excerpt from the card by Moondream, licensed other.
Identity and Version
- Repository
- moondream/moondream3.1-9B-A2B
- Publisher
- Moondream
- Task
- Image and text to text
- Modality
- Image and text
- Library
- Not stated by the source
- Parameters
- 9.3B parameters
- Languages
- Not stated by the source
- Revision
- 1da27e6c2d2f9118de14560df456881dc62d91f3
- First published
- 2026-06-30
- Last updated
- 2026-07-08
Files and Weights
4 files, 10.5 GB in total. The weights are 1 file totalling 10.5 GB in safetensors.
Weights1 file · 10.5 GB
Configuration1 file · 38 B
Documentation1 file · 2.5 KB
Repository1 file · 1.5 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model.safetensors | Weights | 10.5 GB | 26dde16af961 |
| config.json | Configuration | 38 B | — |
| README.md | Documentation | 2.5 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- other
- Access
- Open weights, no gate
- Download size
- 10.5 GB
Download from Moondream
Released by Moondream through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 10.5 GB |
| 16-bit | 18.5 GB |
| 8-bit | 9.3 GB |
| 4-bit | 4.6 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About moondream3.1-9B-A2B
How much GPU memory does moondream3.1-9B-A2B need?
About 22.3 GB at 16-bit and 5.6 GB at 4-bit: the weights (9.3B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run moondream3.1-9B-A2B on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
What license is moondream3.1-9B-A2B released under?
other, as its publisher declares it. Read the license text before commercial use.
Similar Models
Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance. Qwen3.5 represents a significant leap forward, integrating breakthroughs in multimodal learning, architectural efficiency, reinforcement learning scale, and global accessibility to empower developers and enterprises with unprecedented capability and efficiency. For more details, please refer to our blog post Qwen3.5. WMT24++: a harder subset of WMT24 after difficulty labeling and rebalancing; we report the averaged scores on 55 languages using XCOMET-XXL. Empty cells (--) indicate scores not yet available or not applicable. Empty cells (--) indicate scores not…
This repo quantizes the model using data-free quantization technique. As of 2026-02-25, make sure your system has cuda12.8 installed. Then, create a fresh Python environment (e.g. python3.12 venv) and run: Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance. Qwen3.5 represents a significant leap forward, integrating breakthroughs in multimodal learning, architectural efficiency, reinforcement learning scale, and global accessibility to empower developers and enterprises with unprecedented capability and efficiency. For more details, please refer to our blog post Qwen3.5. WMT24++: a harder subset of WMT24 after…
Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date. This generation delivers comprehensive upgrades across the board: superior text understanding & generation, deeper visual perception & reasoning, extended context length, enhanced spatial and video dynamics comprehension, and stronger agent interaction capabilities. Available in Dense and MoE architectures that scale from edge to cloud, with Instruct and reasoning‑enhanced Thinking editions for flexible, on‑demand deployment. Text Understanding on par with pure LLMs: Seamless text–vision fusion for lossless, unified comprehension. 1. Interleaved-MRoPE: Full‑frequency allocation over time, width, and height…
Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date. This generation delivers comprehensive upgrades across the board: superior text understanding & generation, deeper visual perception & reasoning, extended context length, enhanced spatial and video dynamics comprehension, and stronger agent interaction capabilities. Available in Dense and MoE architectures that scale from edge to cloud, with Instruct and reasoning‑enhanced Thinking editions for flexible, on‑demand deployment. Text Understanding on par with pure LLMs: Seamless text–vision fusion for lossless, unified comprehension. 1. Interleaved-MRoPE: Full‑frequency allocation over time, width, and height…
Anonymous release for the ICLR 2027 submission MemGUI-RL: Reinforcement Learning for Proactive Context Management in Long-Horizon Mobile GUI Agents. Project page: https://memgui-rl-anonymous.github.io/ MemGUI-8B-RL is MemGUI-8B-SFT (Qwen3-VL-8B-Instruct supervised on MemGUI-3K) post-trained for 100 optimizer steps with FARPO (Folding-Aware Reward-decoupled Policy Optimization, span-to-step ratio rho = 9). The policy speaks the ConAct (Context-as-Action) interface of MemGUI-Agent: every response contains a folding directive for its own history, an optional memory operation and the next GUI action. The checkpoint is a standard Qwen3VLForConditionalGeneration model (weights in bf16, ~17.5 GB).…