AImodelsComparison.com 1,086 open · 200 API
Menu
Models Best of Trending Can I run it? Methodology About
1,086 models · 1,086 open weights · 200 API-only · refreshed nightly

Open-weights models

Models whose weights are published on Hugging Face — download them and run them on your own hardware. Many are also served by hosted providers; those carry an "Also via API" badge with live prices.

Type
All 1,286 Open weights (self-host) 1,086 API-only 200
Task
All text-generation 461 text-to-image 150 image-text-to-text 101 ASR 83 sentence-similarity 56 translation 46 feature-extraction 44 text-classification 43 image-classification 36 token-classification 17 text-to-speech 17 audio-classification 9 image-segmentation 8 object-detection 7 summarization 6 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
$
1,086 models
#ModelParamsContextCommercial use30dMin VRAM
501 nllb-200-distilled-1.3B facebook · translation Open weights 1.3 B 1 K ✗ cc-by-nc-4.0 485 K from 3.6 GB
502 Qwen3-8B-FP8 Qwen · text-generation Open weights 8.2 B 41 K ✓ apache-2.0 482 K from 12.1 GB
503 sanoTTS ampixa · text-to-speech Open weights — — gpl-3.0 482 K from 0.5 GB
504 pythia-14m EleutherAI · text-generation Open weights 10 M 2 K ✓ apache-2.0 481 K from 0.5 GB
505 german-sentiment-bert oliverguhr · text-classification Open weights 109 M 512 ✓ mit 477 K from 1.0 GB
506 sat-3l-sm segment-any-text · token-classification Open weights 210 M 514 ✓ mit 477 K from 1.5 GB
507 multilingual-e5-small Xenova · feature-extraction Open weights 120 M 512 unknown 476 K from 0.8 GB
508 conv-bert-base YituTech · feature-extraction Open weights — 512 unknown 476 K —
509 sentence-bert-base-ja-mean-tokens-v2 sonoisa · feature-extraction Open weights 110 M 512 cc-by-sa-4.0 474 K from 1.0 GB
510 wav2vec2-large-xlsr-japanese-hiragana vumichien · ASR Open weights 320 M — ✓ apache-2.0 472 K from 1.9 GB
511 mask2former-swin-large-ade-semantic facebook · image-segmentation Open weights 220 M — other 471 K from 1.5 GB
512 faster-whisper-tiny.en Systran · ASR Open weights — — ✓ mit 469 K —
513 Qwen3.8-2B-Distill-GGUF empero-ai · text-generation Open weights — — ✓ apache-2.0 468 K from 1.9 GB
514 wav2vec2-large-xlsr-53-basque stefan-it · ASR Open weights 320 M — ✓ apache-2.0 468 K from 1.9 GB
515 JiRackDeltaNet_27b CMSManhattan · text-generation Open weights 27.3 B 262 K ✓ mit 468 K from 16.5 GB
516 GigaAM-v3 ai-sage · ASR Open weights — — ✓ mit 467 K —
517 gpt-neo-125m EleutherAI · text-generation Open weights 150 M 2 K ✓ mit 467 K from 1.1 GB
518 gte-base-en-v1.5 Alibaba-NLP · sentence-similarity Open weights 137 M 8 K ✓ apache-2.0 467 K from 1.1 GB
519 KR-SBERT-V40K-klueNLI-augSTS snunlp · sentence-similarity Open weights — 512 unknown 467 K —
520 xlm-emo-t MilaNLProc · text-classification Open weights — 512 unknown 465 K —
521 Laguna-S-2.1 poolside · text-generation Open weights Also via API 117.6 B 1.0 M openmdw-1.1 464 K from 276.8 GB
522 Hermes-3-Llama-3.1-8B NousResearch · text-generation Open weights 8.0 B 131 K ⚠ llama3 464 K from 19.4 GB
523 Qwen3-VL-30B-A3B-Instruct-AWQ QuantTrio · text-generation Open weights 31.1 B — ✓ apache-2.0 462 K from 24.8 GB
524 Llama-3.1-405B-FP8 meta-llama · text-generation · gated Open weights 405.9 B — ⚠ llama3.1 462 K from 597.3 GB
525 pplx-embed-v1-0.6b perplexity-ai · feature-extraction Open weights 600 M 33 K ✓ mit 461 K from 3.2 GB
526 Ornith-1.5-9B ornith-ai · text-generation Open weights 9.7 B — ✓ mit 461 K from 23.2 GB
527 Qwen2.5-7B Qwen · text-generation Open weights 7.6 B 131 K ✓ apache-2.0 460 K from 18.4 GB
528 Qwen-Image-2.1-FP8 unsloth · text-to-image Open weights 7.1 B — other 460 K from 46.7 GB
529 Qwen3.6-35B-A3B-abliterated-v4 Bahushruth · text-generation Open weights 34.7 B 262 K ✓ apache-2.0 460 K from 82.0 GB
530 mimi kyutai · feature-extraction Open weights 100 M 8 K ✓ cc-by-4.0 458 K from 0.9 GB
531 Qwen3-Coder-30B-A3B-Instruct-GGUF unsloth · text-generation Open weights — — ✓ apache-2.0 456 K from 12.9 GB
532 MedCPT-Query-Encoder ncbi · feature-extraction Open weights 110 M 512 other 454 K from 1.0 GB
533 GLM-4.7-Flash-AWQ QuantTrio · text-generation Open weights 31.2 B 203 K ✓ mit 454 K from 26.9 GB
534 Apodex-1.1-mini-GGUF abenzerps · text-generation Open weights — — ✓ apache-2.0 452 K from 10.2 GB
535 Qwen2.5-72B-Instruct-AWQ Qwen · text-generation Open weights 73.0 B 33 K other 447 K from 57.2 GB
536 larger_clap_music_and_speech laion · feature-extraction Open weights — — ✓ apache-2.0 447 K —
537 audiobox-aesthetics facebook · audio-classification Open weights 100 M — ✓ cc-by-4.0 445 K from 1.0 GB
538 indobert-base-p1 indobenchmark · feature-extraction Open weights — 512 ✓ mit 444 K —
539 segformer_b2_clothes mattmdjaga · image-segmentation Open weights 30 M — other 442 K from 0.6 GB
540 Qwen2.5-Coder-7B-Instruct-AWQ Orion-zhen · text-generation Open weights 7.6 B 33 K ✓ apache-2.0 441 K from 7.9 GB
541 bge-m3-spa-law-qa littlejohn-ai · sentence-similarity · gated Open weights 570 M — ✓ apache-2.0 440 K from 3.1 GB
542 canine-c google · feature-extraction Open weights 130 M 16 K ✓ apache-2.0 439 K from 1.1 GB
543 Ornith-1.5-397B-NVFP4 ornith-ai · text-generation Open weights 397.0 B — ✓ mit 439 K from 411.4 GB
544 MiniMax-M2.7 MiniMaxAI · text-generation Open weights Also via API 228.7 B 205 K other 438 K from 288.0 GB
545 Llama-2-7b-hf meta-llama · text-generation · gated Open weights 6.7 B — ⚠ llama2 436 K from 16.3 GB
546 Qwen3-TTS-12Hz-1.7B-VoiceDesign Qwen · text-to-speech Open weights 1.9 B — ✓ apache-2.0 435 K from 5.8 GB
547 vit_base_patch16_224.augreg2_in21k_ft_in1k timm · image-classification Open weights 90 M — ✓ apache-2.0 435 K from 0.9 GB
548 phi-2 microsoft · text-generation Open weights 2.8 B 2 K ✓ mit 434 K from 7.0 GB
549 lambda unslothai · feature-extraction Open weights — — unknown 434 K from 0.5 GB
550 Qwen2.5-Coder-3B-Instruct Qwen · text-generation Open weights 3.1 B 33 K other 434 K from 7.8 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.