Best GGUF models for llama.cpp and Ollama
GGUF is the format llama.cpp uses, and it is supported by Ollama and LM Studio. GGUF models can run on a CPU or GPU, and they are often quantized to reduce memory usage. These are the most popular GGUF builds on Hugging Face.
Last updated 3 November 2026 · refreshed after every Hugging Face sync · 15 models
| # | Model | Size | Memory | License | Downloads (30 days) | Likes |
|---|---|---|---|---|---|---|
| 1 | Qwen3.8-Flash-Next-GSQ-RCO-GGUF ISTA-DASLab | – · GGUF | – | Apache 2.0 | 7.4M | 1.1K |
| 2 | Ornith-1.5-9B-GGUF ornith-ai | 9B · GGUF | ~6 GB | MIT | 5.4M | 625 |
| 3 | Ternary-Bonsai-2-27B-gguf prism-ml | 27B · GGUF | ~17 GB | Apache 2.0 | 1.8M | 2.9K |
| 4 | Ornith-1.5-35B-A3B-GGUF ornith-ai | 35B · GGUF | ~21 GB | MIT | 4M | 571 |
| 5 | Qwen3.8-27B-GSQ-RCO-GGUF ISTA-DASLab | 27B · GGUF | ~17 GB | Apache 2.0 | 1.3M | 2.5K |
| 6 | Ornith-1.0-9B-GGUF ornith-ai | 9B · GGUF | ~6 GB | MIT | 2.4M | 699 |
| 7 | DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF ISTA-DASLab | – · GGUF | – | MIT | 2.4M | 565 |
| 8 | Ornith-1.0-35B-GGUF ornith-ai | 35B · GGUF | ~21 GB | MIT | 1.5M | 1.1K |
| 9 | Qwen3.8-Flash-Next-GGUF unsloth | – · GGUF | – | Custom (other) | 1.1M | 1.2K |
| 10 | Qwen3-Coder-30B-A3B-Instruct-GGUF unsloth | 30B · GGUF | ~18 GB | Apache 2.0 | 1.1M | 1.2K |
| 11 | Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF ukisai | 27B · GGUF | ~17 GB | Custom (other) | 1.6M | 478 |
| 12 | deepseek-v4-gguf antirez | – · GGUF | – | MIT | 1.1M | 495 |
| 13 | Qwen3.8-Flash-Next-GSQ-RCO-Coder-GGUF ISTA-DASLab | – · GGUF | – | Apache 2.0 | 889.3K | 545 |
| 14 | Qwen3.8-35B-A3B-Distill-GGUF empero-ai | 35B · GGUF | ~21 GB | Apache 2.0 | 1.3M | 257 |
| 15 | GLM-5.3-Flash-GGUF unsloth | – · GGUF | – | MIT | 828.7K | 514 |
How we rank
Language and vision-language models in GGUF format. We include models seen on Hugging Face in the last 14 days with at least 10,000 downloads in the last 30 days. Score = log10(30-day downloads) + ½ × log10(likes + 1), so popularity leads and likes break near-ties. We keep one entry per model name and leave out adult and uncensored builds. Memory is estimated from parameter count and the published file format (16-bit ≈ 2 bytes per parameter, 4-bit ≈ 0.5) plus 20% headroom; long contexts need more. A quantised copy of a bigger model may also fit — check the model page.