AImodelsComparison.com 1,213 open · 210 API
Menu
Models Best of Trending Can I run it? Methodology MCP server About
1,423 models · 1,213 open weights · 210 API-only · refreshed nightly

Models that run on 2 × 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.

Type
All 1,423 Open weights (self-host) 1,213 API-only 210
Task
All text-generation 735 text-to-image 165 image-text-to-text 110 ASR 82 sentence-similarity 63 text-classification 50 translation 50 feature-extraction 46 image-classification 41 text-to-speech 23 token-classification 22 summarization 9 audio-classification 9 image-segmentation 9 object-detection 7 audio-text-to-text 1 any-to-any 1
Runs on
Any RTX 3060 · 12 GB RTX 4070 · 16 GB RTX 4090 · 24 GB 2 × 24 GB Mac M3 · 24 GB Mac M4 Max · 64 GB
Licence
Size
$
871 models
#ModelParamsContextCommercial use30dOn 2 × 24 GB
501 Meta-Llama-3.1-8B-Instruct-AWQ-INT4 hugging-quants · text-generation Open weights 8.0 B 131 K ⚠ llama3.1 321 K 8.0 GB
502 Phi-3-mini-4k-instruct microsoft · text-generation Open weights 3.8 B 4 K ✓ mit 321 K 9.5 GB
503 Ornith-1.0-9B-AWQ-INT4 cyankiwi · text-generation Open weights 9.4 B — ✓ mit 317 K 11.8 GB
504 Spark-X2.5-4B-GGUF XHToken · text-generation Open weights 4.0 B — ✓ apache-2.0 315 K 4.0 GB
505 regnety_032.ra_in1k timm · image-classification Open weights 20 M — ✓ apache-2.0 314 K 0.6 GB
506 Qwen3.8-27B-MTP-NVFP4 sakamakismile · text-generation Open weights 27.4 B — ✓ apache-2.0 313 K 27.2 GB
507 opus-mt-fr-en Helsinki-NLP · translation Open weights 80 M 512 ✓ apache-2.0 312 K 0.8 GB
508 Qwen3.6-27B-NVFP4 nvidia · text-generation Open weights 18.2 B — ✓ apache-2.0 312 K 27.3 GB
509 macbert4csc-base-chinese shibing624 · text-generation Open weights 100 M 512 ✓ apache-2.0 312 K 1.0 GB
510 phishing-email-detection-distilbert_v2.4.1 cybersectony · text-classification Open weights 70 M 512 ✓ apache-2.0 311 K 0.8 GB
511 flint-GGUF unsloth · text-generation Open weights 39.8 B 262 K ✓ apache-2.0 308 K 19.0 GB
512 mms-lid-1024 facebook · audio-classification Open weights 967 M — ✗ cc-by-nc-4.0 308 K 4.9 GB
513 gpt-oss-120b-GGUF ggml-org · text-generation Open weights 120.0 B 131 K ✓ apache-2.0 306 K 19.4 GB
514 Qwen3-1.7B PrimeIntellect · text-generation Open weights 2.0 B 41 K ✓ apache-2.0 306 K 5.3 GB
515 prokbert-mini-promoter neuralbioinfo · text-classification Open weights 20 M 1 K ✗ cc-by-nc-4.0 302 K 0.6 GB
516 vit_small_patch16_224.augreg_in21k_ft_in1k timm · image-classification Open weights 20 M — ✓ apache-2.0 302 K 0.6 GB
517 pythia-14m EleutherAI · text-generation Open weights 10 M 2 K ✓ apache-2.0 299 K 0.5 GB
518 GLM-4.7-Flash-GGUF unsloth · text-generation Open weights — — ✓ mit 296 K 13.0 GB
519 nllb-200-3.3B facebook · translation Open weights 3.3 B 1 K ✗ cc-by-nc-4.0 296 K 8.3 GB
520 Apertus-8B-Instruct-2509 swiss-ai · text-generation Open weights 8.1 B 66 K ✓ apache-2.0 295 K 19.4 GB
521 JiRackUltra_14b CMSManhattan · text-generation Open weights 14.8 B 131 K ✓ mit 292 K 9.1 GB
522 voice-gender-classifier JaesungHuh · audio-classification Open weights 20 M — ✓ mit 291 K 0.6 GB
523 Qwen2.5-3B-Instruct-GGUF Qwen · text-generation Open weights — — other 291 K 2.0 GB
524 Qwen2.5-7B-Instruct-GGUF bartowski · text-generation Open weights — — ✓ apache-2.0 288 K 3.6 GB
525 xflux_text_encoders XLabs-AI · text-generation Open weights 4.8 B — ✓ apache-2.0 288 K 11.7 GB
526 resnet50.fb_swsl_ig1b_ft_in1k timm · image-classification Open weights 26 M — ✗ cc-by-nc-4.0 288 K 0.6 GB
527 animagine-xl-3.1 cagliostrolab · text-to-image Open weights 2.6 B — ⚠ openrail++ 286 K 16.1 GB
528 segformer-b0-finetuned-ade-512-512 nvidia · image-segmentation Open weights <0.1 M — other 286 K 0.5 GB
529 gpt2-mini erwanf · text-generation Open weights 39 M — ✓ mit 285 K 0.7 GB
530 controlnet-openpose-sdxl-1.0 xinsir · text-to-image Open weights 1.3 B — ✓ apache-2.0 285 K 6.2 GB
531 LFM3-1.7B-GGUF LiquidAI · text-generation Open weights 1.7 B — other 283 K 1.9 GB
532 koelectra-small-v3-nsmc daekeun-ml · text-classification Open weights 10 M 512 ✓ mit 283 K 0.6 GB
533 OpenMed-NER-PharmaDetect-SuperClinical-434M OpenMed · token-classification Open weights 430 M 512 ✓ apache-2.0 279 K 1.5 GB
534 vit_base_patch8_224.augreg2_in21k_ft_in1k timm · image-classification Open weights 90 M — ✓ apache-2.0 279 K 0.9 GB
535 Qwen3-14B-FP8 Qwen · text-generation Open weights 14.8 B 41 K ✓ apache-2.0 277 K 20.7 GB
536 llm-jp-4-33b-thinking-gguf llm-jp · text-generation Open weights 33.0 B — ✓ apache-2.0 277 K 27.6 GB
537 Qwen2-7B-Instruct Qwen · text-generation Open weights 7.6 B 33 K ✓ apache-2.0 276 K 18.4 GB
538 vit_tiny_patch16_224.augreg_in21k_ft_in1k timm · image-classification Open weights 10 M — ✓ apache-2.0 276 K 0.5 GB
539 phi-4 microsoft · text-generation Open weights Also via API 14.7 B 16 K ✓ mit 276 K 34.9 GB
540 convnext_tiny.fb_in22k_ft_in1k timm · image-classification Open weights 29 M — ✓ apache-2.0 275 K 0.6 GB
541 LFM2.5-230M-Base LiquidAI · text-generation Open weights 230 M 128 K other 275 K 1.0 GB
542 Qwen3.8-27B-NVFP4-MTP-GGUF esatapedico · text-generation Open weights 27.0 B — ✓ apache-2.0 275 K 41.0 GB
543 IP-Adapter-FaceID h94 · text-to-image Open weights — — unknown 274 K 1.5 GB
544 Ternary-Bonsai-27B-gguf prism-ml · text-generation Open weights — — ✓ apache-2.0 274 K 1.2 GB
545 vit_tiny_r_s16_p8_224.augreg_in21k timm · image-classification Open weights 10 M — ✓ apache-2.0 274 K 0.5 GB
546 LFM2.5-24B-A2B-GGUF LiquidAI · text-generation Open weights 24.0 B 131 K other 273 K 19.0 GB
547 GLM-4.7-Flash-AWQ QuantTrio · text-generation Open weights 31.2 B 203 K ✓ mit 272 K 26.9 GB
548 edgenext_small.usi_in1k timm · image-classification Open weights 10 M — ✓ mit 271 K 0.5 GB
549 table-transformer-detection microsoft · object-detection Open weights 30 M 1 K ✓ mit 270 K 0.6 GB
550 CodonTransformer adibvafa · token-classification Open weights 90 M 4 K ✓ apache-2.0 270 K 0.9 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.