181K models tracked · refreshed 30 days ago
Track the models shaping open-source AI
Daily download history, growth metrics and rankings for open-source AI models — built from our own observations, so you can see how a model got here, not just where it stands today.
Free to use · No account required
- Models trackedModels currently monitored by OpenModelStats' daily ingestion.
- 181K
- ObservationsHistorical snapshots recorded over time — the core of OpenModelStats' trend data.
- 3.7M
- PublishersDistinct organizations and owners behind tracked models.
- 46.3K
- Last refreshMost recent successful data refresh from the source.
- 30 days ago
Trending now
View all →- 1DeepSeek-V4.1-Flash-1017deepseek-aiImage Text to Text1,975—
- 2flux-3-image-klein-comfyblack-forest-labsImage To Image914—
- 3Qwen-Image-2.1-GGUFabenzerpsText to Image914—
- 4Qwen-Image-2.1-Uncensored-GGUFabenzerpsText to Image602—
- 5Muse-Spark-Open-816B-A42Bmeta-modelsImage Text to Text582—
- 6flux-3-image-kleinblack-forest-labsImage To Image543—
- 7MOSS-Image-SFTOpenMOSS-TeamText to Image506—
- 8TeleOCRStarDoc-AIImage Text to Text49679.6%(up)
- 9drama-3fishaudioText to Speech494—
- 10MOSS-Image-BaseOpenMOSS-TeamText to Image449—
Most downloaded
View all →- 1all-MiniLM-L6-v2sentence-transformersSentence Similarity198.6M1.6%(up)
- 2ms-marco-MiniLM-L6-v2cross-encoderText Ranking73.1M3.8%(up)
- 3bge-small-en-v1.5BAAIEmbeddings55.4M3.4%(up)
- 4paraphrase-multilingual-MiniLM-L12-v2sentence-transformersSentence Similarity41.9M1.9%(up)
- 5electra-base-discriminatorgoogle38.9M0.4%(up)
- 6bge-m3BAAISentence Similarity30.4M3.0%(up)
- 7bert-base-uncasedgoogle-bertFill Mask28.9M0.4%(up)
- 8Qwen3-0.6BQwenText Generation26.7M4.3%(up)
- 9MiniMax-H3Comfy-Org25.8M16.4%(up)
- 10t5-smallgoogle-t5Translation24.1M2.2%(up)
Fastest growing
View all →Downloads gained in the last 7 tracked days
- 1all-MiniLM-L6-v2sentence-transformersSentence Similarity+59.9M1.6%(up)
- 2ms-marco-MiniLM-L6-v2cross-encoderText Ranking+21.2M3.8%(up)
- 3bge-small-en-v1.5BAAIEmbeddings+15.5M3.4%(up)
- 4bert-base-uncasedgoogle-bertFill Mask+11.9M0.4%(up)
- 5paraphrase-multilingual-MiniLM-L12-v2sentence-transformersSentence Similarity+11.3M1.9%(up)
- 6Qwen3-0.6BQwenText Generation+9.2M4.3%(up)
- 7bge-m3BAAISentence Similarity+8.9M3.0%(up)
- 8mobilenetv3_small_100.lamb_in1ktimmImage Classification+5.9M0.5%(up)
- 9t5-smallgoogle-t5Translation+5.8M2.2%(up)
- 10chronos-2amazonTime Series Forecasting+5.7M3.4%(up)
New & rising
View all →- 1MiniMax-H3Comfy-Org25.8M16.4%(up)
- 2Qwen-Image-2.1Comfy-Org10.5M—
- 3Qwen3.8-27BQwenImage Text to Text7.8M19.2%(up)
- 4Qwen3.8-Flash-Next-GSQ-RCO-GGUFISTA-DASLabImage Text to Text6.4M—
- 5GLM-5.3-Flashzai-orgImage Text to Text6.4M87.0%(up)
- 6Qwen3.8-27B-GGUFunsloth5.9M15.0%(up)
- 7YuE2-GGUFServeurperso5.4M151.4%(up)
- 8Ornith-1.5-9B-GGUFornith-aiText Generation5.1M26.7%(up)
- 9MiniMax-H3MiniMaxAIImage Text To Video5M13.2%(up)
- 10Qwen3.8-27B-FP8QwenImage Text to Text4.3M14.7%(up)
Explore by task
All tasks →Text Generation
52.8K models · 440.3M dl/30d
Sentence Similarity
2,608 models · 395.4M dl/30d
Image Text to Text
14.7K models · 293.8M dl/30d
Speech Recognition
8,731 models · 180.1M dl/30d
Embeddings
4,994 models · 154.2M dl/30d
Fill Mask
1,471 models · 110.2M dl/30d
Text Ranking
453 models · 95.6M dl/30d
Image Classification
3,402 models · 69.8M dl/30d
Top publishers
All publishers →- 1sentence-transformers292.6M
- 2Qwen255.3M
- 3BAAI134.9M
- 4google120.8M
- 5cross-encoder86.7M
- 6Comfy-Org86.3M
- 7facebook58.4M
- 8unsloth55.5M
- 9openai54.9M
- 10timm47.2M
What OpenModelStats tracks
Historical snapshots
Every day, OpenModelStats records what it observes about each tracked model — downloads, likes and rank positions — building a history that did not exist anywhere else.
How observation works →Growth & velocity
Interval gains are derived only from changes in cumulative counters between OpenModelStats observations. When a number cannot be calculated reliably, we show nothing at all.
How metrics are computed →Independent rankings
Rankings compare tracked models against each other and are labeled as such. Accuracy matters more than impressive-sounding numbers.
See current rankings →