Skip to content
AI Compute Radar
IndexCollected 3 h ago

Community versions

Versions made by other people

Most of what people download is not a maker's original model. It is a version someone else trained a little further and shared. This list ranks those versions by downloads and says, for your card, whether each one runs.

87 versions of 36 models, measured.

  1. Minimax-h3_Singularity

    made by WarmBloodAban · a version of MiniMax-H3 · Video

    • 1.3 million downloads in the last 30 days
    • File: 11.8 GB

    On the RTX 4090RunsLargest part held at once: 15.7 GB.

    See it on Hugging Face ↗ (External link)
  2. Cyber-Tiel-Coder-35B-A3B-MTP

    made by peculiar-ragdoll · a version of Ornith 1.5 35B-A3B · Text

    • 971,000 downloads in the last 30 days
    • File: 21.0 GB

    On the RTX 4090Spills into RAMNeeds about 22.6 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  3. NVIDIA-Nemotron-3.5-Bolt-E4B

    made by nvidia · a version of Gemma 4 E4B · Text

    • 732,000 downloads in the last 30 days
    • File: 4.7 GB

    On the RTX 4090Runs easilyNeeds about 5.9 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  4. FastVideo-FastH3-Comfy

    made by FastVideo · a version of MiniMax-H3 · Video

    • 644,000 downloads in the last 30 days
    • File: 22.1 GB

    On the RTX 4090Too bigLargest part held at once: 22.1 GB.

    See it on Hugging Face ↗ (External link)
  5. minimax-h3-fused-turbo-int8-convrot

    made by MATLOWAI · a version of MiniMax-H3 · Video

    • 388,000 downloads in the last 30 days
    • File: 21.0 GB

    On the RTX 4090Barely fitsLargest part held at once: 21.0 GB.

    See it on Hugging Face ↗ (External link)
  6. s1-mini

    made by superwhisper · a version of Qwen3 0.6B · Text

    • 233,000 downloads in the last 30 days
    • File: 0.5 GB

    On the RTX 4090Runs easilyNeeds about 2.2 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  7. Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated

    made by huihui-ai · a version of Qwen3-Coder 30B-A3B · Text

    • 140,000 downloads in the last 30 days
    • File: 17.3 GB

    On the RTX 4090Barely fitsNeeds about 19.4 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  8. Huihui-Qwen3-8B-abliterated-v2

    made by huihui-ai · a version of Qwen3 8B · Text

    • 95,000 downloads in the last 30 days
    • File: 4.7 GB

    On the RTX 4090Runs easilyNeeds about 6.8 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  9. Huihui-Qwen3-14B-abliterated-v2

    made by huihui-ai · a version of Qwen3 14B · Text

    • 91,000 downloads in the last 30 days
    • File: 8.4 GB

    On the RTX 4090Runs easilyNeeds about 10.8 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  10. Artemis-31B-v1.2

    made by TheDrummer · a version of Gemma 4 31B · Text

    • 87,000 downloads in the last 30 days
    • File: 18.2 GB

    On the RTX 4090Barely fitsNeeds about 21.0 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  11. gemma-4-31B-it-uncensored-heretic

    made by llmfan46 · a version of Gemma 4 31B · Text

    • 84,000 downloads in the last 30 days
    • File: 17.4 GB

    On the RTX 4090Barely fitsNeeds about 20.2 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  12. Orion-26B-A4B-v1.1

    made by TheDrummer · a version of Gemma 4 26B-A4B · Text

    • 66,000 downloads in the last 30 days
    • File: 16.7 GB

    On the RTX 4090RunsNeeds about 18.4 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  13. Qwen3.8-27B-Humanlike-Chat

    made by LessThanThreeAI · a version of Qwen3.8 27B · Text

    • 62,000 downloads in the last 30 days
    • File: 15.4 GB

    On the RTX 4090RunsNeeds about 17.2 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  14. Huihui-Qwen3-Coder-Next-abliterated

    made by huihui-ai · a version of Qwen3-Coder Next · Text

    • 60,000 downloads in the last 30 days
    • File: 45.2 GB

    On the RTX 4090Spills into RAMNeeds about 47.6 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  15. Huihui-Qwen3.8-27B-abliterated

    made by huihui-ai · a version of Qwen3.8 27B · Text

    • 53,000 downloads in the last 30 days
    • File: 15.7 GB

    On the RTX 4090RunsNeeds about 17.5 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  16. Hermes-4-14B

    made by NousResearch · a version of Qwen3 14B · Text

    • 52,000 downloads in the last 30 days
    • File: 8.4 GB

    On the RTX 4090Runs easilyNeeds about 10.8 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  17. Cydonia-24B-v4.3

    made by TheDrummer · a version of Mistral Small 3.2 · Text

    • 52,000 downloads in the last 30 days
    • File: 13.3 GB

    On the RTX 4090RunsNeeds about 15.8 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  18. Llama-3.3-70B-Instruct-abliterated

    made by huihui-ai · a version of Llama 3.3 70B · Text

    • 49,000 downloads in the last 30 days
    • File: 39.6 GB

    On the RTX 4090Spills into RAMNeeds about 44.1 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  19. gemma-4-26B-A4B-it-heretic

    made by coder3101 · a version of Gemma 4 26B-A4B · Text

    • 42,000 downloads in the last 30 days
    • File: 15.6 GB

    On the RTX 4090RunsNeeds about 17.3 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  20. gemma-4-31B-it-heretic

    made by coder3101 · a version of Gemma 4 31B · Text

    • 38,000 downloads in the last 30 days
    • File: 17.4 GB

    On the RTX 4090Barely fitsNeeds about 20.2 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  21. Huihui-GLM-4.7-Flash-abliterated

    made by huihui-ai · a version of GLM-4.7-Flash · Text

    • 33,000 downloads in the last 30 days
    • File: 16.9 GB

    On the RTX 4090RunsNeeds about 18.7 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  22. L3.3-70B-Euryale-v2.3

    made by Sao10K · a version of Llama 3.3 70B · Text

    • 32,000 downloads in the last 30 days
    • File: 39.6 GB

    On the RTX 4090Spills into RAMNeeds about 44.1 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  23. Huihui-Qwen3-4B-Instruct-2507-abliterated

    made by huihui-ai · a version of Qwen3 4B · Text

    • 31,000 downloads in the last 30 days
    • File: 2.3 GB

    On the RTX 4090Runs easilyNeeds about 4.3 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  24. MOSS-Image-SFT-Turbo-INT8-ConvRot

    made by toxicdog · a version of MOSS-Image · Image

    • 31,000 downloads in the last 30 days
    • File: 5.4 GB

    On the RTX 4090Runs easilyLargest part held at once: 7.3 GB.

    See it on Hugging Face ↗ (External link)
  25. Huihui-Ornith-1.5-35B-A3B-abliterated

    made by huihui-ai · a version of Ornith 1.5 35B-A3B · Text

    • 27,000 downloads in the last 30 days
    • File: 20.2 GB

    On the RTX 4090Spills into RAMNeeds about 21.8 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  26. GLM-4.7-Flash-REAP-23B-A3B

    made by cerebras · a version of GLM-4.7-Flash · Text

    • 23,000 downloads in the last 30 days
    • File: 13.1 GB

    On the RTX 4090Runs easilyNeeds about 14.8 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  27. Nemotron-Orchestrator-8B

    made by nvidia · a version of Qwen3 8B · Text

    • 23,000 downloads in the last 30 days
    • File: 4.7 GB

    On the RTX 4090Runs easilyNeeds about 6.8 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  28. clef

    made by Cloudflare · a version of Qwen3.8 27B · Text

    • 20,000 downloads in the last 30 days
    • File: 18.2 GB

    On the RTX 4090Barely fitsNeeds about 20.1 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  29. Huihui-gemma-4-12B-it-abliterated

    made by huihui-ai · a version of Gemma 4 12B · Text

    • 16,000 downloads in the last 30 days
    • File: 6.9 GB

    On the RTX 4090Runs easilyNeeds about 8.4 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)
  30. Winnow-12B

    made by EldanRing · a version of Gemma 4 12B · Text

    • 15,000 downloads in the last 30 days
    • File: 11.8 GB

    On the RTX 4090Runs easilyNeeds about 13.8 GB with room for a few pages of chat.

    See it on Hugging Face ↗ (External link)

How this list is made

Sorted by downloads and nothing else. We have not tried any of these versions and do not say which one is better.

Every file size is read from the files on Hugging Face. For text models the memory for the chat is computed from the version's own settings; for image and video models the word is judged on the largest part that has to be in memory at once.

Left out: small add-on files that only work together with the original, repacked copies of the original, other editions by the original's own maker, versions marked for adults only, and anything that cannot be measured.

Last read on 19 November 2026The method →All models →