# moondream2 by Vik Korrapati: Open-Weight Model Source: https://savrn.com/models/moondream2 Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve moondream2 (1.9B 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 | 3.9 GB | 4.6 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 8-bit | 1.9 GB | 2.3 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 4-bit | 1.0 GB | 1.2 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/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](https://savrn.com/ai-index/pricing/gpus), read Oct 23, 2026. [moondream2 on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/moondream2/gpus) ## SAVRN's Notes on moondream2 Image and text in, text out, from 1.9 billion parameters. At 16-bit the weights are 3.9 GB and the model needs 4.6 GB to run; 8-bit brings that to 2.3 GB and 4-bit to 1.2 GB. The cheapest rental we list is a single MI300X with 192 GB at $1.85 an hour on-demand, so the question is not whether moondream2 fits but how many copies you stack on one card before paying for a second. Apache 2.0 permits commercial use, modification and redistribution, provided the license, copyright notices and any NOTICE file travel with it and significant changes are stated. Before committing, note that the publisher calls this repository the previous generation, with Moondream 3 in preview, and updates it often enough to recommend pinning a revision. We have no context length on record for it, so confirm the per-call input budget with the publisher. SAVRN Research, 2026-09-18 ## Model Card By Vik Korrapati, published under apache-2.0, revision 5d6c926f44e2. This repository contains the latest version of Moondream 2, our previous generation model. The latest version of Moondream is[Moondream 3 (Preview)](https://huggingface.co/moondream/moondream3-preview). Moondream is a small vision language model designed to run efficiently everywhere. [Website](https://moondream.ai/) / [Demo](https://moondream.ai/playground) / [GitHub](https://github.com/vikhyat/moondream) This repository contains the latest (2025-06-21) release of Moondream 2, as well as [historical releases](https://huggingface.co/vikhyatk/moondream2/blob/main/versions.txt). The model is updated frequently, so we recommend specifying a revision as shown below if you're using it in a production application. ### Usage [Read the full model card (486 words)](https://savrn.com/models/moondream2/card) ## Configuration Architecture HfMoondream Stored precision bfloat16 Model type moondream1 ## Identity and Version Repository vikhyatk/moondream2 Publisher Vik Korrapati Task Image and text to text Modality Image and text Library transformers Parameters 1.9B parameters Languages Not stated by the source Revision 5d6c926f44e26b07957b0dd315bbedcb4c17a5fe First published 2024-03-04 Last updated 2026-09-23 ## Files and Weights 30 files, 3.9 GB in total. The weights are 1 file totalling 3.9 GB in safetensors. Weights1 file · 3.9 GB Configuration21 files · 108.0 KB Tokenizer4 files · 3.4 MB Documentation1 file · 4.4 KB Other2 files · 173 B Repository1 file · 1.6 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model.safetensors | Weights | 3.9 GB | 70a7d94c0c83 | | added_tokens.json | Configuration | 1.1 KB | — | | config.json | Configuration | 277 B | — | | config.py | Configuration | 2.6 KB | — | | configuration_moondream.py | Configuration | 3.3 KB | — | | fourier_features.py | Configuration | 558 B | — | | generation_config.json | Configuration | 69 B | — | | handler.py | Configuration | 2.1 KB | — | | hf_moondream.py | Configuration | 6.6 KB | — | | image_crops.py | Configuration | 8.1 KB | — | | layers.py | Configuration | 4.6 KB | — | | lora.py | Configuration | 2.3 KB | — | | moondream.py | Configuration | 35.5 KB | — | | region.py | Configuration | 4.5 KB | — | | region_model.py | Configuration | 1.3 KB | — | | rope.py | Configuration | 1.6 KB | — | | special_tokens_map.json | Configuration | 99 B | — | | text.py | Configuration | 6.9 KB | — | | utils.py | Configuration | 1.4 KB | — | | vision.py | Configuration | 5.1 KB | — | | vision_encoder.py | Configuration | 10.2 KB | — | | weights.py | Configuration | 9.7 KB | — | | README.md | Documentation | 4.4 KB | — | | requirements.txt | Other | 42 B | — | | versions.txt | Other | 131 B | — | | .gitattributes | Repository | 1.6 KB | — | | merges.txt | Tokenizer | 456.3 KB | — | | tokenizer.json | Tokenizer | 2.1 MB | — | | tokenizer_config.json | Tokenizer | 7.3 KB | — | | vocab.json | Tokenizer | 798.2 KB | — | ## License and Download License apache-2.0 Access Open weights, no gate Download size 3.9 GB [Download from Vik Korrapati](https://huggingface.co/vikhyatk/moondream2) Released by Vik Korrapati through its official repository on Hugging Face. [Read the license](https://www.apache.org/licenses/LICENSE-2.0). ## Evaluations Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical. | Benchmark | Conditions | Result | Reported by | Revision | Date | | --- | --- | --- | --- | --- | --- | | tiiuae/PBench | Task averageMetric averageSetup Detection-only model combined with SAM2 to convert boxes to segmentation masks.Comparison conditions not established | 44.7 | [Community Evals](https://huggingface.co/vikhyatk/moondream2) Reported by a third party | Evaluated revision not stated | 2026-05-11 | ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 3.9 GB | | 16-bit | 3.9 GB | | 8-bit | 1.9 GB | | 4-bit | 1.0 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Questions About moondream2 ### How much GPU memory does moondream2 need? About 4.6 GB at 16-bit and 1.2 GB at 4-bit: the weights (1.9B parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run moondream2 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. ### Can I use moondream2 commercially? Yes. moondream2 is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors. ## Similar Models Model · Image and text to text ### [Qwen3-VL-2B-Instruct](https://savrn.com/models/qwen3-vl-2b-instruct) [Qwen](https://savrn.com/model-publishers/qwen) 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… Open weights apache-2.0 2.1B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-vl-2b-instruct) Model · Image and text to text ### [qwen3-vl-2b-medical-pmcvqa-sft-merged](https://savrn.com/models/qwen3-vl-2b-medical-pmcvqa-sft-merged) [HarisshRagav](https://savrn.com/model-publishers/gl3mon) This qwen3vl model was trained 2x faster with Unsloth and Huggingface's TRL library. Open weights apache-2.0 2.1B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-vl-2b-medical-pmcvqa-sft-merged) Model · Image and text to text ### [Qwen2-VL-2B-Instruct](https://savrn.com/models/qwen2-vl-2b-instruct) [Qwen](https://savrn.com/model-publishers/qwen) We're excited to unveil Qwen2-VL, the latest iteration of our Qwen-VL model, representing nearly a year of innovation. SoTA understanding of images of various resolution & ratio: Qwen2-VL achieves state-of-the-art performance on visual understanding benchmarks, including MathVista, DocVQA, RealWorldQA, MTVQA, etc. Understanding videos of 20min+: Qwen2-VL can understand videos over 20 minutes for high-quality video-based question answering, dialog, content creation, etc. Agent that can operate your mobiles, robots, etc.: with the abilities of complex reasoning and decision making, Qwen2-VL can be integrated with devices like mobile phones, robots, etc., for automatic operation based on… Open weights apache-2.0 2.2B parameters 32,768 tokens transformers [View model](https://savrn.com/models/qwen2-vl-2b-instruct) Model · Image and text to text ### [Qari-OCR-v0.3-VL-2B-Instruct](https://savrn.com/models/qari-ocr-v0-3-vl-2b-instruct) [Network for Advancing Modern ArabicNLP & AI](https://savrn.com/model-publishers/namaa-space) QARI-OCR v0.3 is a specialized vision-language model fine-tuned for Arabic Optical Character Recognition with a focus on structural document understanding. - Built on Qwen2-VL-2B-Instruct, this model excels at preserving document layouts, HTML tags, and formatting while transcribing Arabic text. - It is described in detail in the paper QARI-OCR: High-Fidelity Arabic Text Recognition through Multimodal Large Language Model Adaptation. While QARI v0.2 achieves better raw text accuracy (CER: 0.061), QARI v0.3 excels in: - HTML/Markdown structure preservation - Document layout understanding - Handwritten text recognition (initial capabilities) - 5x faster training than v0.2 You can load this… Open weights apache-2.0 2.2B parameters 32,768 tokens transformers [View model](https://savrn.com/models/qari-ocr-v0-3-vl-2b-instruct) Model · Image and text to text ### [Qwen3.5-2B-Base-heretic](https://savrn.com/models/qwen3-5-2b-base-heretic) [Roumiany](https://savrn.com/model-publishers/aminroumiany) 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. If you find our work helpful, feel free to give us a cite. Open weights apache-2.0 2.2B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-5-2b-base-heretic) Model · Image and text to text ### [Qwen3.5-2B-Vi-SFT-VLM-v2.1](https://savrn.com/models/qwen3-5-2b-vi-sft-vlm-v2-1) [Pham Phuc](https://savrn.com/model-publishers/phuc-hugigface) This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019). Open weights 2.2B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-5-2b-vi-sft-vlm-v2-1) ## Vik Korrapati [All models and datasets](https://savrn.com/model-publishers/vikhyatk) ## Versions - [5d6c926f44e2](https://savrn.com/models/moondream2/versions/5d6c926f44e2) · current 2026-10-23 - [6b714b26eea5](https://savrn.com/models/moondream2/versions/6b714b26eea5) 2026-09-23 ## Explore More - [All image and text to text models](https://savrn.com/models/tasks/image-and-text-to-text) - [All models under apache-2.0](https://savrn.com/models/licenses/apache-2-0) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-09-24. - [Hugging Face record](https://huggingface.co/vikhyatk/moondream2) - [How the hub is built](https://savrn.com/model-hub/methodology)