Models › Runs on my machine
Runs on my machine
For running it yourself on a 16 GB card or a 64 GB Mac (unified memory).
- Fastest first token112msDeepSeek-V4.1-FlashTogether AI via huggingface.co, 2026-10-25
- Cheapest with tool calls$0.010per 1M inQwen3-Embedding-8BDeepInfra via huggingface.co, 2026-10-25
- Most fine-tuned 2025 base2.6kfine-tunesQwen3-4B-Instruct-2507Hugging Face base-model tags, 2026-10-25
Text models whose INT4 weights fit the memory you pick, with a GGUF build on the Hub, released since 2024. 64 and 128 mean a Mac's unified memory. Sorted by 30-day downloads. Showing models that fit 16 GB at INT4.
The floor
Every view leaves out test fixtures, models whose size didn't parse, and closed models (unless you ask for them). $0 endpoints show as a free tier and never count as cheapest; classifiers and guards never enter a speed sort. How the rank works.
267 models match, first 60 shown. Data observed 2026-10-25 06:54 UTC, Hub snapshot 2026-10-25. Hover a figure for its source and date.
| # | Evidence | GGUF · MLX | ||||||
|---|---|---|---|---|---|---|---|---|
| 1 | Qwen3-0.6BQwen | 752M | ~0.5 GB | GGUF · MLX | Apache 2.0 | |||
| 2 | Qwen3-4BQwen | 4.02B | ~2.4 GB | GGUF · MLX | Apache 2.0 | |||
| 3 | Qwen3-14BQwen | 14.8B | ~8.9 GB | GGUF · MLX | Apache 2.0 | |||
| 4 | gemma-4-26B-A4B-itgoogle | 25.8B 4B active | ~16 GB | GGUF · MLX | Apache 2.0 | |||
| 5 | Qwen3-8BQwen | 8.19B | ~4.9 GB | GGUF · MLX | Apache 2.0 | |||
| 6 | Qwen3.5-4BQwen | 4.66B | ~2.8 GB | GGUF · MLX | Apache 2.0 | |||
| 7 | Qwen3.5-9BQwen | 9.65B | ~5.8 GB | GGUF · MLX | Apache 2.0 | |||
| 8 | Qwen2.5-7B-InstructQwen | 7.62B | ~4.6 GB | GGUF | Apache 2.0 | |||
| 9 | gpt-oss-20bopenai | 20.9B | ~13 GB | GGUF · MLX | Apache 2.0 | |||
| 10 | Qwen2.5-0.5B-InstructQwen | 494M | ~0.3 GB | GGUF · MLX | Apache 2.0 | |||
| 11 | Qwen3-VL-8B-InstructQwen | 8.77B | ~5.3 GB | GGUF · MLX | Apache 2.0 | |||
| 12 | Qwen2.5-VL-7B-InstructQwen | 8.29B | ~5.0 GB | GGUF · MLX | Apache 2.0 | |||
| 13 | Llama-3.1-8B-Instructmeta-llama | 8.03B | ~4.8 GB | GGUF · MLX | Llama 3.1 Community Lic… | |||
| 14 | Qwen2.5-1.5B-InstructQwen | 1.54B | ~0.9 GB | GGUF | Apache 2.0 | |||
| 15 | Qwen3-1.7BQwen | 2.03B | ~1.2 GB | GGUF · MLX | Apache 2.0 | |||
| 16 | gemma-4-E4B-itgoogle | 8.00B | ~4.8 GB | GGUF · MLX | Apache 2.0 | |||
| 17 | Qwen3-VL-4B-InstructQwen | 4.44B | ~2.7 GB | GGUF · MLX | Apache 2.0 | |||
| 18 | Qwen2.5-3B-InstructQwen | 3.09B | ~1.9 GB | GGUF · MLX | Qwen License (qwen-rese… | |||
| 19 | Llama-3.2-1B-Instructmeta-llama | 1.24B | ~0.7 GB | GGUF | Llama 3.2 Community Lic… | |||
| 20 | Kimi-K3-DSparkRadixArk | 2.25B | ~1.3 GB | GGUF | Not declared | |||
| 21 | Qwen3.5-0.8BQwen | 873M | ~0.5 GB | GGUF · MLX | Apache 2.0 | |||
| 22 | Qwen3.5-2BQwen | 2.27B | ~1.4 GB | GGUF · MLX | Apache 2.0 | |||
| 23 | Florence-2-basemicrosoft | 232M | ~0.1 GB | GGUF | MIT | |||
| 24 | gemma-4-E2B-itgoogle | 5.12B | ~3.1 GB | GGUF · MLX | Apache 2.0 | |||
| 25 | Qwen3-4B-Instruct-2507Qwen | 4.02B | ~2.4 GB | GGUF · MLX | Apache 2.0 | |||
| 26 | Unlimited-OCRbaidu | 3.34B | ~2.0 GB | GGUF · MLX | MIT | |||
| 27 | moondream2vikhyatk | 1.93B | ~1.2 GB | GGUF | Apache 2.0 | |||
| 28 | DeepSeek-OCRdeepseek-ai | 3.34B | ~2.0 GB | GGUF · MLX | MIT | |||
| 29 | SmolLM2-135MHuggingFaceTB | 135M | ~0.1 GB | GGUF | Apache 2.0 | |||
| 30 | Qwen3-VL-2B-InstructQwen | 2.13B | ~1.3 GB | GGUF · MLX | Apache 2.0 | |||
| 31 | Qwen2.5-Coder-14B-InstructQwen | 14.8B | ~8.9 GB | GGUF · MLX | Apache 2.0 | |||
| 32 | gemma-4-12B-itgoogle | 12.0B | ~7.2 GB | GGUF · MLX | Apache 2.0 | |||
| 33 | Qwen2.5-VL-3B-InstructQwen | 3.75B | ~2.3 GB | GGUF · MLX | Qwen License (qwen-rese… | |||
| 34 | Qwen2.5-Coder-7B-InstructQwen | 7.62B | ~4.6 GB | GGUF · MLX | Apache 2.0 | |||
| 35 | surya-ocr-2datalab-to | 686M | ~0.4 GB | GGUF · MLX | OpenRAIL | |||
| 36 | chandra-ocr-2datalab-to | 5.30B | ~3.2 GB | GGUF · MLX | OpenRAIL | |||
| 37 | DeepSeek-R1-Distill-Qwen-1.5Bdeepseek-ai | 1.78B | ~1.1 GB | GGUF · MLX | MIT | |||
| 38 | MiniCPM5-2Bopenbmb | 2.52B | ~1.5 GB | GGUF · MLX | Apache 2.0 | |||
| 39 | Qwen2.5-0.5BQwen | 494M | ~0.3 GB | GGUF · MLX | Apache 2.0 | |||
| 40 | Qwen2-VL-2B-InstructQwen | 2.21B | ~1.3 GB | GGUF | Apache 2.0 | |||
| 41 | Llama-3.2-3B-Instructmeta-llama | 3.21B | ~1.9 GB | GGUF · MLX | Llama 3.2 Community Lic… | |||
| 42 | gemma-3-4b-itgoogle | 4.30B | ~2.6 GB | GGUF | Gemma Terms of Use | |||
| 43 | Llama-3.2-1B-Instructunsloth | 1.24B | ~0.7 GB | GGUF | Llama 3.2 Community Lic… | |||
| 44 | gemma-3-1b-itgoogle | 1000M | ~0.6 GB | GGUF · MLX | Gemma Terms of Use | |||
| 45 | HunyuanOCRtencent | 1.12B | ~0.7 GB | GGUF · MLX | tencent-hunyuan-communi… | |||
| 46 | Qwen2.5-14B-InstructQwen | 14.8B | ~8.9 GB | GGUF · MLX | Apache 2.0 | |||
| 47 | Qwen3-8B-BaseQwen | 8.19B | ~4.9 GB | GGUF · MLX | Apache 2.0 | |||
| 48 | Qwen3.5-4B-BaseQwen | 4.66B | ~2.8 GB | GGUF · MLX | Apache 2.0 | |||
| 49 | SmolLM3-3BHuggingFaceTB | 3.08B | ~1.8 GB | GGUF · MLX | Apache 2.0 | |||
| 50 | Qwen3-4B-Thinking-2507Qwen | 4.02B | ~2.4 GB | GGUF · MLX | Apache 2.0 | |||
| 51 | Qwen2.5-1.5BQwen | 1.54B | ~0.9 GB | GGUF · MLX | Apache 2.0 | |||
| 52 | Qwen3-0.6B-BaseQwen | 596M | ~0.4 GB | GGUF · MLX | Apache 2.0 | |||
| 53 | diffusiongemma-26B-A4B-itgoogle | 25.8B 4B active | ~16 GB | GGUF · MLX | Apache 2.0 | |||
| 54 | Qwen2-VL-7B-InstructQwen | 8.29B | ~5.0 GB | GGUF · MLX | Apache 2.0 | |||
| 55 | granite-4.1-3bibm-granite | 3.40B | ~2.0 GB | GGUF · MLX | Apache 2.0 | |||
| 56 | Llama-3.1-8Bmeta-llama | 8.03B | ~4.8 GB | GGUF | Llama 3.1 Community Lic… | |||
| 57 | Cosmos-Reason2-2Bnvidia | 2.44B | ~1.5 GB | GGUF | NVIDIA Open Model Licen… | |||
| 58 | Qwen3-4B-BaseQwen | 4.02B | ~2.4 GB | GGUF · MLX | Apache 2.0 | |||
| 59 | LightOnOCR-2-1Blightonai | 1.01B | ~0.6 GB | GGUF | Apache 2.0 | |||
| 60 | Qwen3-1.7B-BaseQwen | 1.72B | ~1.0 GB | GGUF · MLX | Apache 2.0 |
Evidence marks, left to right: coding, tools, reasoning, long docs, vision, chat. measured on a third-party board, author-reported, structural, none yet. Each mark links to its job from the model page. INT4 is an estimate of the weights alone (method).