selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

Alternatives to ZwZ2-4B

120 open-weight models that do the same job as ZwZ2-4B (4.5B vision), ranked by how close they are in task and size, then by popularity and how recent they are. Licence and VRAM are shown for each.

the original
inclusionAI logo

ZwZ2-4B

inclusionAI·Vision

≈4.2 GB

ZwZ2-4B is the Qwen3.5-4B-based model in Zooming-without-Zooming-v2 (ZwZ2). It learns to connect visual evidence across regions through comparison, aggregation, matching, and multi-step reasoning. Training uses full images; inference requires no zooming tools.

4.5B256K ctx Unknown licence
1.1KOct 2026gguf
--vram
any
--gpus
--quant
  1. moondream logo

    moondream3-preview

    moondream·Vision

    ≈7.1 GB

    Moondream 3 (Preview) is an vision language model with a mixture-of-experts architecture (9B total parameters, 2B active). This model makes no compromises, delivering state-of-the-art visual reasoning while still retaining our efficient and deployment-friendly ethos.

    9.3BCustom licence
    36.7KSept 2025
    • Same category
    • Larger (9.3B vs 4.5B)
    vs
  2. docling-project logo

    DeskForge-Gemma4-E4B

    docling-project·Vision

    ≈5.9 GB

    DeskForge-Gemma4-E4B is Gemma 4 E4B (instruction-tuned) fine-tuned on 200K grounding examples from DeskForge-1M. Given a desktop screenshot and a natural-language instruction, it answers with the next action as a pyautogui call whose coordinates are fractions of the screen.

    7.9B128K ctx Apache 2.0
    446Oct 2026gguf
    • Same category
    • Similar size (7.9B vs 4.5B)
    • Commercial use OK
    vs
  3. FreedomIntelligence logo

    HuatuoGPT-3-9B

    FreedomIntelligence·Vision

    ≈7.1 GB

    HuatuoGPT-3-9B is a medical LLM built on Qwen3.5-9B with One-stage Policy Optimization (OnePO). OnePO adapts language models to medicine in a single reinforcement-learning stage, without preceding domain-specific supervised fine-tuning.

    9.4B256K ctx Apache 2.0
    2.1KSept 2026gguf
    • Same category
    • Larger (9.4B vs 4.5B)
    • Commercial use OK
    vs
  4. baidu logo

    Qianfan-VL-3B

    baidu·Vision

    ≈3.0 GB

    Qianfan-VL is a series of general-purpose multimodal large language models enhanced for enterprise-level multimodal applications. The models offer deep optimization for high-frequency scenarios in industrial deployment while maintaining strong general capabilities.

    3.7B32K ctx Custom licence
    273Sept 2025
    • Same category
    • Similar size (3.7B vs 4.5B)
    vs
  5. google logo

    paligemma2-3b-mix-224

    google·Vision

    ≈3.2 GB

    Transformers PaliGemma 2 3B weights fine-tuned on a mixture of academic tasks using 224×224 input images. PaliGemma 2 mix checkpoints are fine-tuned on a diverse set of tasks and are ready to use out of the box while pt checkpoints are pre-trained and intended for further fine-tuning.

    3BGemma Terms of Use
    16.3KNov 2024
    • Same category
    • Similar size (3B vs 4.5B)
    vs
  6. google logo

    t5gemma-2-1b-1b

    google·Vision

    ≈2.0 GB

    T5Gemma is a family of lightweight yet powerful encoder-decoder research models from Google, built by adapting pretrained decoder-only Gemma models into encoder-decoder ones.

    2.1B32K ctx Gemma Terms of Use
    33.3KOct 2025
    • Same category
    • Smaller (2.1B vs 4.5B)
    vs
  7. datalab-to logo

    lift

    datalab-to·Vision

    ≈7.2 GB

    lift is a structured extraction model from Datalab that pulls structured JSON out of PDFs and images. Pass any JSON schema and lift returns a JSON object matching it, using schema-constrained decoding to guarantee valid, well-typed output.

    9.7B256K ctx OpenRAIL
    4.6KJun 2026gguf
    • Same category
    • Larger (9.7B vs 4.5B)
    vs
  8. google logo≈3.2 GB

    Transformers PaliGemma 2 3B weights, fine-tuned with 448*448 input images on the DOCCI dataset.

    3BGemma Terms of Use
    14.9KNov 2024
    • Same category
    • Similar size (3B vs 4.5B)
    vs
  9. inclusionAI logo

    UI-Venus-1.5-2B

    inclusionAI·Vision

    ≈2.8 GB

    This repository contains the UI-Venus model from the report UI-Venus-1.5 Technical Report. UI-Venus 1.5 is a unified, end-to-end GUI Agent designed for robust real-world applications. The model family includes two dense (2B/8B) and one MoE (30B-A3B) variants to meet various downstream scenarios.

    2.4B256K ctx Apache 2.0
    3.6KFeb 2026gguf
    • Same category
    • Similar size (2.4B vs 4.5B)
    • Commercial use OK
    vs
  10. microsoft logo

    Fara1.5-9B

    microsoft·Vision

    ≈7.1 GB

    Fara1.5-9B is a multimodal computer use agent (CUA) for web browsers, from Microsoft Research AI Frontiers. It observes the browser through screenshots and acts on the user's behalf by emitting structured tool calls — click, type, scroll, visit URL, web search, and so on — to complete tasks…

    9.4B256K ctx MIT
    4.3KMay 2026gguf
    • Same category
    • Larger (9.4B vs 4.5B)
    • Commercial use OK
    vs
  11. tencent logo

    UI-Mate-9B

    tencent·Vision

    ≈7.1 GB

    Show the workflow once. Let the agent adapt it to the task at hand.

    9.4B256K ctx Apache 2.0
    2.2KAug 2026gguf
    • Same category
    • Larger (9.4B vs 4.5B)
    • Commercial use OK
    vs
  12. internlm logo

    Intern-S1-mini

    internlm·Vision

    ≈6.7 GB

    💻Github Repo • 🤗Model Collections • 📜Technical Report • 🏠Project Page

    8.5B64K ctx Apache 2.0
    13.2KAug 2025gguf
    • Same category
    • Similar size (8.5B vs 4.5B)
    • Commercial use OK
    vs
  13. Salesforce logo

    blip2-flan-t5-xl

    Salesforce·Vision

    ≈2.9 GB

    BLIP-2 model, leveraging Flan T5-xl (a large language model). It was introduced in the paper BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models by Li et al. and first released in this repository.

    3.9B512 ctx MIT
    19.1KFeb 2023
    • Same category
    • Similar size (3.9B vs 4.5B)
    • Commercial use OK
    vs
  14. google logo

    paligemma-3b-mix-224

    google·Vision

    ≈2.4 GB

    Transformers PaliGemma 3B weights, fine-tuned with 224*224 input images and 256 token input/output text sequences on a mixture of downstream academic datasets. The models are available in float32, bfloat16 and float16 format for research purposes only.

    2.9BGemma Terms of Use
    72KMay 2024
    • Same category
    • Similar size (2.9B vs 4.5B)
    vs
  15. HuggingFaceTB logo

    SmolVLM2-2.2B-Instruct

    HuggingFaceTB·Vision

    ≈2.0 GB

    SmolVLM2-2.2B is a lightweight multimodal model designed to analyze video content. The model processes videos, images, and text inputs to generate text outputs - whether answering questions about media files, comparing visual content, or transcribing text from images.

    2.2B8K ctx Apache 2.0
    74.7KFeb 2025gguf
    • Same category
    • Smaller (2.2B vs 4.5B)
    • Commercial use OK
    vs
  16. apple logo

    LensVLM-9B

    apple·Vision

    ≈7.1 GB

    LensVLM is a 9B Vision Language Model (VLM) that scans compressed images of text, then selectively expands only the relevant pages to their uncompressed form via learned tools.

    9.4B256K ctx Apple Machine Learning Research License
    917Sept 2026gguf
    • Same category
    • Larger (9.4B vs 4.5B)
    vs
  17. nvidia logo

    Cosmos-Reason1-7B

    nvidia·Vision

    ≈5.8 GB

    NVIDIA Cosmos Reason – an open, customizable, 7B-parameter reasoning vision language model (VLM) for physical AI and robotics - enables robots and vision AI agents to reason like humans, using prior knowledge, physics understanding and common sense to understand and act in the real world.

    8.3B125K ctx NVIDIA Open Model License
    15.8KApr 2025gguf
    • Same category
    • Similar size (8.3B vs 4.5B)
    • Commercial use OK
    vs
  18. ByteDance logo

    Sa2VA-4B

    ByteDance·Vision

    ≈3.1 GB

    Sa2VA is an MLLM capable of question answering, visual prompt understanding, and dense object segmentation at both image and video levels. It achieves comparable performance to SOTA MLLMs Qwen2-VL and InternVL2.5 on question-answering benchmarks.

    3.9B32K ctx Apache 2.0
    470Jan 2025
    • Same category
    • Similar size (3.9B vs 4.5B)
    • Commercial use OK
    vs
  19. microsoft logo

    Fara-7B

    microsoft·Vision

    ≈5.8 GB

    Description: Fara-7B is Microsoft's first agentic small language model (SLM) designed specifically for computer use. With only 7 billion parameters, Fara-7B is an ultra-compact Computer Use Agent (CUA) that achieves state-of-the-art performance within its size class and is competitive with larger…

    8.3B125K ctx MIT
    3.2KOct 2025gguf
    • Same category
    • Similar size (8.3B vs 4.5B)
    • Commercial use OK
    vs
  20. inclusionAI logo

    UI-Venus-1.5-8B

    inclusionAI·Vision

    ≈6.8 GB

    This repository contains the UI-Venus model from the report UI-Venus-1.5 Technical Report. UI-Venus 1.5 is a unified, end-to-end GUI Agent designed for robust real-world applications. The model family includes two dense (2B/8B) and one MoE (30B-A3B) variants to meet various downstream scenarios.

    8.8B256K ctx Apache 2.0
    2.3KFeb 2026gguf
    • Same category
    • Similar size (8.8B vs 4.5B)
    • Commercial use OK
    vs
  21. llava-hf logo

    llava-v1.6-mistral-7b-hf

    llava-hf·Vision

    ≈6.0 GB

    The LLaVA-NeXT model was proposed in LLaVA-NeXT: Improved reasoning, OCR, and world knowledge by Haotian Liu, Chunyuan Li, Yuheng Li, Bo Li, Yuanhan Zhang, Sheng Shen, Yong Jae Lee.

    7.6BApache 2.0
    146KFeb 2024
    • Same category
    • Similar size (7.6B vs 4.5B)
    • Commercial use OK
    vs
  22. XiaomiMiMo logo

    MiMo-Embodied-7B

    XiaomiMiMo·Vision

    ≈6.5 GB

    MiMo-Embodied, a powerful cross-embodied vision-language model that shows state-of-the-art performance in both autonomous driving and embodied AI tasks, the first open-source VLM that integrates these two critical areas, significantly enhancing understanding and reasoning in dynamic physical…

    8.3B125K ctx MIT
    2.6KNov 2025gguf
    • Same category
    • Similar size (8.3B vs 4.5B)
    • Commercial use OK
    vs
  23. XiaomiMiMo logo

    MiMo-VL-7B-RL-2508

    XiaomiMiMo·Vision

    ≈6.5 GB

    We're excited to announce improvements to our MiMo-VL (MiMo-VL-7B-RL-2508 and MiMo-VL-7B-SFT-2508), featuring enhanced performance across multiple benchmarks, improved thinking control capabilities, and better user experience.

    8.3B125K ctx MIT
    5.8KAug 2025gguf
    • Same category
    • Similar size (8.3B vs 4.5B)
    • Commercial use OK
    vs
  24. tencent logo

    WeMM-Video-2B

    tencent·Vision

    ≈2.8 GB

    WeMM-Video-2B is the dense 2B checkpoint in the WeMM-Video family. WeMM-Video is a Chinese–English bilingual family for joint audio-visual understanding and reasoning. It accepts text, images, audio, and video and generates text responses.

    3.3B256K ctx Apache 2.0
    11Nov 2026gguf
    • Same category
    • Similar size (3.3B vs 4.5B)
    • Commercial use OK
    • Newer release
    vs