1,423 models · 1,213 open weights · 210 API-only · refreshed nightly
Models that run on Mac M3 · 24 GB
Every model in the catalog — open weights you can self-host and API-only models — with licence or API price, estimated VRAM and daily-tracked downloads. Filters update the URL — share any view.
| # | Model | Params | Context | Commercial use | 30d | On Mac M3 · 24 GB |
|---|---|---|---|---|---|---|
| 601 | pythia-160m-deduped | 210 M | 2 K | ✓ apache-2.0 | 145 K | 0.9 GB |
| 602 | Qwopus3.6-27B-Fusion-GGUF | — | — | other | 145 K | 15.4 GB |
| 603 | Granite-4.2-3B-BE-GGUF | 3.0 B | — | ✓ apache-2.0 | 144 K | 2.6 GB |
| 604 | Nanbeige_Nanbeige4.3-33B-GGUF | 33.0 B | — | ✓ apache-2.0 | 140 K | 16.5 GB |
| 605 | DeepSeek-R1-Distill-Qwen-7B-GGUF | 7.0 B | 131 K | unknown | 139 K | 4.6 GB |
| 606 | typhoon2.5-qwen3-4b | 4.0 B | 262 K | ✓ apache-2.0 | 139 K | 10.0 GB |
| 607 | JevK5 | 4.2 B | 262 K | ✓ apache-2.0 | 138 K | 10.4 GB |
| 608 | needle3 | — | 8 K | ✓ apache-2.0 | 136 K | 0.8 GB |
| 609 | Wan2.2-I2V_General-NSFW-LoRA | — | — | unknown | 136 K | 1.8 GB |
| 610 | Qwen-Image-GGUF | 20.4 B | — | ✓ apache-2.0 | 136 K | 12.1 GB |
| 611 | P-Q-Image-2.1 | 7.1 B | — | other | 134 K | 3.8 GB |
| 612 | MOSS-Image-SFT-GGUF | — | — | ✓ apache-2.0 | 134 K | 3.6 GB |
| 613 | NVIDIA-Nemotron-3.5-Bolt-E4B-4GB | 4.0 B | — | other | 132 K | 7.9 GB |
| 614 | Llama-3.2-1B-Instruct-GGUF | — | — | ⚠ llama3.2 | 132 K | 1.2 GB |
| 615 | lcm-lora-sdv1-5 | — | — | ⚠ openrail++ | 122 K | 0.6 GB |
| 616 | Krea-2-Turbo-GGUF | — | — | other | 114 K | 5.9 GB |
| 617 | Qwen-Image-2.1-GGUF | — | — | unknown | 109 K | 3.3 GB |
| 618 | Flux2-Klein-9B-True-V2 | — | — | other | 109 K | 6.8 GB |
| 619 | janku-v5-nsfw-trained-noobai-rou-wei-illustrious-xl-v50-sdxl | 2.6 B | — | other | 108 K | 8.5 GB |
| 620 | Qwen-Image-GGUF | — | — | ✓ apache-2.0 | 107 K | 0.8 GB |
| 621 | z-image-turbo-nsfw-v2-GGUF | — | — | ✓ apache-2.0 | 105 K | 8.4 GB |
| 622 | HunyuanDiT-v1.1-Diffusers-Distilled | 1.5 B | — | other | 101 K | 16.7 GB |
| 623 | opus-mt-tc-big-ko-en | 210 M | 1 K | ✓ cc-by-4.0 | 95 K | 1.0 GB |
| 624 | ideogram-4-gguf | 9.3 B | — | other | 94 K | 8.1 GB |
| 625 | FLUX.1-schnell-GGUF | 11.9 B | — | ✓ apache-2.0 | 91 K | 6.7 GB |
| 626 | amanatsu-illustrious-v11-sdxl | 2.6 B | — | other | 90 K | 8.5 GB |
| 627 | Qwen-Image-2.1-Turbo-Quantized | — | — | other | 90 K | 5.0 GB |
| 628 | noobai-XL-Vpred-1.0 | 2.6 B | — | other | 88 K | 16.3 GB |
| 629 | Anima-2.9B | — | — | other | 86 K | 10.3 GB |
| 630 | T5-base-10K-summarization | 220 M | — | unknown | 84 K | 1.5 GB |
| 631 | noobai-XL-1.0 | 2.6 B | — | other | 83 K | 16.1 GB |
| 632 | FLUX.2-klein-4B-mflux-4bit | — | — | ✓ apache-2.0 | 80 K | 5.6 GB |
| 633 | InstantID | — | — | ✓ apache-2.0 | 79 K | 3.3 GB |
| 634 | diving-illustrious-real-asian-v50-sdxl | 2.6 B | — | other | 79 K | 8.5 GB |
| 635 | LFM3-1.7B | 1.7 B | 262 K | other | 79 K | 4.6 GB |
| 636 | Chroma-GGUF | — | — | ✓ apache-2.0 | 78 K | 6.0 GB |
| 637 | dvine82-xl | 2.6 B | — | unknown | 77 K | 8.5 GB |
| 638 | Index-Translate-9B-GGUF | 9.0 B | — | ✓ apache-2.0 | 76 K | 6.2 GB |
| 639 | controlnet-canny-sdxl-1.0 | 1.3 B | — | ✓ apache-2.0 | 75 K | 6.2 GB |
| 640 | Qwen-Image-2.1-Turbo-GGUF | — | — | other | 75 K | 3.2 GB |
| 641 | sd-xl-alternative-bases | — | — | ⚠ openrail++ | 73 K | 15.9 GB |
| 642 | Z-Image-Turbo-FP8 | — | — | ✓ apache-2.0 | 72 K | 14.0 GB |
| 643 | novaAnimeXL_ilV140 | 2.6 B | — | unknown | 72 K | 16.1 GB |
| 644 | Qwen-Image-gguf | — | — | ✓ apache-2.0 | 71 K | 8.3 GB |
| 645 | controlnet-openpose-sdxl-1.0 | — | — | other | 70 K | 6.9 GB |
| 646 | Z-Image-Turbo-GGUF | — | — | ✓ apache-2.0 | 70 K | 4.7 GB |
| 647 | Ministral-3-3B-Instruct-2512 | 3.8 B | — | ✓ apache-2.0 | 66 K | 11.4 GB |
| 648 | stable-diffusion-v1-5-GGUF | — | — | ⚠ creativeml-openrail-m | 66 K | 2.2 GB |
| 649 | dreamshaper-8 | 860 M | — | ⚠ creativeml-openrail-m | 65 K | 9.7 GB |
| 650 | controlnet-depth-sdxl-1.0 | 1.3 B | — | ✓ apache-2.0 | 64 K | 3.4 GB |
VRAM figures are estimates for the smallest available quantization at 8K context — see /methodology. Downloads refresh nightly from the Hugging Face API; API prices (blended 3:1 input:output per 1M tokens) and Artificial Analysis indices from OpenRouter.