formats · tinymodels.co
Tiny safetensors models
Every safetensors model in the catalogue. safetensors is a model format: the figures below are this hub's own measurements, and every download states its sha256.
Browse by task instead: see the catalogue for the task list.
| model | params | size | licence | one fixed input (ms) | |
|---|---|---|---|---|---|
| SmolVLM2-256M-Video-Instruct | 256M | 1.03 GB | apache-2.0 | 2,841.4 | Get |
| Qwen3-0.6B | 752M | 1.50 GB | apache-2.0 | 469.6 | Get |
| Florence-2-base-ft | 232M | 463 MB | mit | 2,124.4 | Get |
| SmolLM2-135M-Instruct | 135M | 269 MB | apache-2.0 | 110.2 | Get |
| videomae-base-finetuned-kinetics | 86.5M | 346 MB | cc-by-nc-4.0 | 214.2 | Get |
| detr-resnet-50 | 41.6M | 167 MB | apache-2.0 | 251.1 | Get |
| Qwen2.5-0.5B-Instruct | 494M | 988 MB | apache-2.0 | 266.8 | Get |
| resnet-50 | 25.6M | 102 MB | apache-2.0 | 21.7 | Get |
| all-MiniLM-L6-v2 | 22.7M | 90.9 MB | apache-2.0 | 5.3 | Get |
| distilbert-base-uncased-finetuned-sst-2-english | 67M | 268 MB | apache-2.0 | 13.1 | Get |
The millisecond figure is one fixed input per task type, measured on this hub's machine with one warm-up run then one timed run. How these numbers were taken; the whole run, every model, is the measured bench dataset.