Models

Popular models from Hugging Face that GridPup can serve, grouped by family and size, with the GPU memory each one needs and the cheapest machine that fits it right now. Open a family to compare its sizes and versions (the original, quantized and ONNX versions need less GPU memory) and the GPUs for each.

2,289 families · catalog updated 2 hours ago

• Qwen3 by QwenText generation · 14 models, 90 versions · ↓ 86.5M0.6B1.7B4B8B14B30B32B235B
  Smallest needs2.9 GB GPU
  From$0.08/hGTX 1660 SUPER
• Qwen2.5 by QwenClassification, Text generation · 7 models, 89 versions · ↓ 45.2M0.5B1.5B3B7B14B32B72B
  Smallest needs1.2 GB GPU
  From$0.08/hGTX 1660 SUPER
• Qwen3.5 by QwenVision + text, Text generation · 9 models, 40 versions · ↓ 33.8M · licence to accept0.8B2B4B9B27B35B122B397B
  Smallest needs3.5 GB GPU
  From$0.10/hRTX 3060
• gemma-4 by googleVision + text, Text generation · 4 models, 25 versions · ↓ 26.5M12B26B31B
  Smallest needs22.6 GB GPU
  From$0.18/hNvidia V100
• Qwen3.6 by QwenVision + text, Text generation · 2 models, 26 versions · ↓ 22.2M27B35B
  Smallest needs25.3 GB GPU
  From$0.34/h2x RTX 5060 Ti
• Qwen3.8 by QwenVision + text, Text generation · 1 model, 25 versions · ↓ 16.9M · licence to accept27B
  Smallest needs15.9 GB GPU
  From$0.14/hRTX A4000
• gpt2 by openai-communityText generation · 1 model, 2 versions · ↓ 16.8M~137M
  Smallest needs1.2 GB GPU
  From$0.08/hGTX 1660 SUPER
• Qwen3-VL by QwenVision + text, Text generation · 10 models, 22 versions · ↓ 15.8M2B4B8B30B32B235B
  Smallest needs8 GB GPU
  From$0.10/hRTX 3060
• Qwen2.5-VL by QwenVision + text, Text generation · 4 models, 11 versions · ↓ 12.6M3B7B32B72B
  Smallest needs6.3 GB GPU
  From$0.10/hRTX 3060
• gpt-oss by openaiText generation · 2 models, 6 versions · ↓ 9.3M20B120B
  Smallest needs16.6 GB GPU
  From$0.18/h2x RTX 3060
• Llama-3.1 by meta-llamaText generation · 4 models, 36 versions · ↓ 8.6M · licence to accept8B70B405B
  Smallest needs8.6 GB GPU
  From$0.09/hRTX 3060
• Llama-3.2 by meta-llamaText generation · 4 models, 18 versions · ↓ 8.2M · licence to accept1B3B
  Smallest needs3.1 GB GPU
  From$0.08/hGTX 1660 SUPER
• GLM-5.3-Flash by zai-orgVision + text, Text generation · 1 model, 4 versions · ↓ 6.7M~321B
  Smallest needs177.3 GB GPU
  From$2.36/h4x Nvidia A40
• pythia by EleutherAIText generation · 19 models · ↓ 6.3M14M31M70M160M410M1B1.4B2.8B6.9B12B
  Smallest needs2.3 GB GPU
  From$0.10/hRTX 3060
• gemma-3 by googleVision + text, Text generation · 5 models, 23 versions · ↓ 6.1M · licence to accept270M1B4B12B27B
  Smallest needs2.6 GB GPU
  From$0.08/hGTX 1660 SUPER
• SmolLM2 by HuggingFaceTBVision + text, Text generation · 3 models, 15 versions · ↓ 5.7M135M360M1.7B
  Smallest needs2 GB GPU
  From$0.08/hGTX 1660 SUPER
• DeepSeek-V3.2-Exp by deepseek-aiText generation · 1 model, 3 versions · ↓ 5.3M~685B
  Smallest needs784.2 GB GPU
  From$44.03/h8x Nvidia H200
• dolphin-2.9.1-yi-1.5 by dphnText generation · 2 models · ↓ 4.7M9B34B
  Smallest needs22.7 GB GPU
  From$0.24/hRTX 3090
• Kimi-K3-DSpark by RadixArkText generation · 1 model · ↓ 4.6M~2.25B
  Smallest needs7.4 GB GPU
  From$0.10/hRTX 3060
• Ornith-1.5 by ornith-aiText generation · 8 models, 10 versions · ↓ 4.3M9B35B397B
  Smallest needs13.2 GB GPU
  From$0.14/hRTX A4000
• DeepSeek-V4-Flash-0731 by deepseek-aiText generation · 1 model, 5 versions · ↓ 4.3M · licence to accept~304B
  Smallest needs121.7 GB GPU
  From$1.38/h8x RTX 3090
• Qwen3.8-Flash-Next by QwenVision + text, Text generation · 1 model, 7 versions · ↓ 3.6M · licence to accept~180B
  Smallest needs121.3 GB GPU
  From$1.38/h8x RTX 3090
• Qwen2-VL by QwenVision + text · 3 models, 6 versions · ↓ 3.5M2B7B72B
  Smallest needs5.8 GB GPU
  From$0.10/hRTX 3060
• MiniCPM5 by openbmbText generation · 9 models, 11 versions · ↓ 3.5M1B2B
  Smallest needs3 GB GPU
  From$0.10/hRTX 3060
• Qwen3-Coder by QwenText generation · 4 models, 11 versions · ↓ 3.4M30B480B
  Smallest needs21.7 GB GPU
  From$0.18/h2x RTX 3060
• NVIDIA-Nemotron-3-Nano by nvidiaText generation · 3 models, 7 versions · ↓ 3.2M4B30B
  Smallest needs9.6 GB GPU
  From$0.10/hRTX 3060
• Unlimited-OCR by baiduVision + text · 1 model, 2 versions · ↓ 2.7M~3.34B
  Smallest needs5.9 GB GPU
  From$0.10/hRTX 3060
• Ornith-1.0 by ornith-aiText generation · 5 models · ↓ 2.7M9B35B397B
  Smallest needs24.3 GB GPU
  From$0.34/h2x RTX 5060 Ti
• JiRackUltra by CMSManhattanText generation · 4 models · ↓ 2.7M1B7B14B32B
  Smallest needs6.4 GB GPU
  From$0.10/hRTX 3060
• Mistral by mistralaiText generation · 3 models, 9 versions · ↓ 2.6M7B
  Smallest needs8 GB GPU
  From$0.10/hRTX 3060

Model details come from Hugging Face and are refreshed daily. GPU memory is GridPup's estimate: the weights, room for an 8,192-token conversation (chat models) and the runtime's own memory. Large models are split across the GPUs of one machine ("4x RTX 4090"). Only machines with a Vast reliability above 95% that download at 400 Mbit/s or more are counted.