Which models can you actually run?
Every leaderboard ranks models you call over an API. This one ranks models you can put on your own machine — by the runtime that loads them, the format they ship in, whether they're quantised, and whether the licence lets you sell what you build.
Read from the index 2026-09-18Two different claims sit behind every bar, and they are kept apart everywhere on this page. Declared means a model card named the runtime. By format means the weights are in a container that runtime reads — which says the file will open, not that the architecture is implemented. A model is counted once if either is true. Nothing declares candle, burn or ort; no one writes a Rust runtime on a model card. That gap is the whole reason this page exists.
Reach, by runtime
declared on the model cardnot declared, but ships a format it reads
Every bar but one is almost entirely light, which is the finding: for most runtimes the evidence is the file, not the card. MLX is the exception — 11,788 cards name it against 471 models shipping an npz, because mlx-community publishes converted weights under a name that says MLX rather than in the format that proves it. It is the one runtime where the card is the better evidence.
The most-used models tflite can load
- Either signal
- 312
- Declared
- 8
- By format
- 310
- Commercial use ok
- 224
- Quantised build
- 14
| Model | Downloads | Params | Quant | Licence | Terms |
|---|---|---|---|---|---|
| openai-community/gpt2 text-generation | 15,584,259 | 137M | — | MIT | commercial ok |
| distilbert/distilgpt2 text-generation | 2,083,338 | 88M | — | Apache-2.0 | commercial ok |
| Synaptics/yolo declared | 350,286 | — | — | AGPL-3.0-only | commercial ok |
| monologg/koelectra-small-v2-distilled-korquad-384 question-answering | 160,705 | 14M | — | — | unknown |
| NN-Dataset/tflite declared | 143,710 | — | — | — | unknown |
| Synaptics/MobileNetV2 declared | 117,220 | — | — | MIT | commercial ok |
| litert-community/Qwen2.5-1.5B-Instruct declared text-generation | 72,707 | 1.5B base model | — | Apache-2.0 | commercial ok |
| litert-community/DeepSeek-R1-Distill-Qwen-1.5B declared text-generation | 62,743 | 1.8B base model | — | MIT | commercial ok |
| richarrrddd/lama_testing.ptl declared | 55,795 | — | — | Apache-2.0 | commercial ok |
| distilbert/distilbert-base-uncased-distilled-squad question-answering | 37,853 | 66M | — | Apache-2.0 | commercial ok |
| desert-ant-labs/emo text-classification | 26,927 | — | — | — | unknown |
| darkmaniac7/bge-small-en-MNN feature-extraction | 26,034 | 33M base model | — | MIT | commercial ok |
| dejanseo/chrome_models | 21,142 | — | — | — | unknown |
| DocWolle/whisper_tflite_models automatic-speech-recognition | 17,441 | — | — | MIT | commercial ok |
| Synaptics/yolov26n_od | 17,279 | — | — | AGPL-3.0-only | commercial ok |
| mouad-zouhdi/sparta-edgetpu-models | 16,648 | — | — | MIT | commercial ok |
| litert-community/Phi-4-mini-instruct text-generation | 13,946 | 3.8B base model | — | MIT | commercial ok |
| ZawShiShawn/gestura-flux2-klein-4b-litert-tflite image-to-image | 12,193 | 3.9B base model | — | Apache-2.0 | commercial ok |
| stabilityai/stable-audio-3-optimized text-to-audio | 10,548 | — | — | — | unknown |
| litert-community/FLUX.2-klein-4B-LiteRT text-to-image | 9,767 | 3.9B base model | — | Apache-2.0 | commercial ok |
| google/magenta-realtime-2 text-to-audio | 8,548 | — | — | CC-BY-4.0 | commercial ok |
| litert-community/Qwen2.5-0.5B-Instruct text-generation | 8,101 | 494M base model | — | Apache-2.0 | commercial ok |
| litert-community/parakeet-tdt-0.6b-v3 | 6,874 | 600M name | — | Apache-2.0 | commercial ok |
| litert-community/whisper-tiny automatic-speech-recognition | 6,307 | 38M base model | — | Apache-2.0 | commercial ok |
| namxike/Mface | 6,225 | — | — | — | unknown |
| desert-ant-labs/shapes image-classification | 5,563 | — | — | — | unknown |
| desert-ant-labs/redact token-classification | 4,365 | — | — | — | unknown |
| argmaxinc/parakeetkit-litert-pro automatic-speech-recognition | 3,772 | — | — | — | unknown |
| litert-community/moonshine-tiny automatic-speech-recognition | 3,721 | 27M base model | — | MIT | commercial ok |
| litert-community/Bonsai-Image-ternary-4B text-to-image | 3,170 | 4.0B name | — | Apache-2.0 | commercial ok |
| qualcomm/Segment-Anything-Model image-segmentation | 3,011 | — | — | Apache-2.0 | commercial ok |
| litert-community/SmolLM-135M-Instruct text-generation | 2,930 | 135M base model | — | Apache-2.0 | commercial ok |
| litert-community/Z-Image-Turbo-LiteRT text-to-image | 2,734 | 6.2B base model | — | Apache-2.0 | commercial ok |
| shadowlilac/gemma-4-e4b-mtp-extraction-effort | 2,712 | — | — | Apache-2.0 | commercial ok |
| Jeremy341/MIRA-AI object-detection | 2,324 | — | — | MIT | commercial ok |
| argmaxinc/whisperkit-litert | 2,247 | — | — | MIT | commercial ok |
| litert-community/embeddinggemma-300m sentence-similarity | 2,238 | 300M name | — | — | conditional |
| litert-community/Gecko-110m-en question-answering | 2,207 | 110M name | — | Apache-2.0 | commercial ok |
| desert-ant-labs/gist text-classification | 2,193 | — | — | — | unknown |
| skytrix/kirexa-mdx-tflite | 2,090 | — | — | MIT | commercial ok |
| NAKSTStudio/yolov8m-chess-piece-detection object-detection | 1,943 | — | — | AGPL-3.0-only | commercial ok |
| litert-community/parakeet-ctc-0.6b automatic-speech-recognition | 1,891 | 609M base model | — | CC-BY-4.0 | commercial ok |
| litert-community/yolox-nano-litert object-detection | 1,745 | — | — | Apache-2.0 | commercial ok |
| devhasni/AnimePixel-Segmentation | 1,662 | — | — | — | unknown |
| nyadla-sys/whisper-tiny.en.tflite automatic-speech-recognition | 1,545 | — | — | MIT | commercial ok |
| litert-community/whisper-acft automatic-speech-recognition | 1,541 | 73M base model | — | Apache-2.0 | commercial ok |
| GabrieleConte/Qwen3.5-0.8B-LiteRT image-text-to-text | 1,486 | 873M base model | — | Apache-2.0 | commercial ok |
| litert-community/Matcha-TTS text-to-speech | 1,485 | — | — | MIT | commercial ok |
| soniqo/VoxCPM2-LiteRT text-to-speech | 1,468 | 2.3B base model | — | Apache-2.0 | commercial ok |
| jegly/noise | 1,307 | — | — | — | unknown |
| litert-community/TinyLlama-1.1B-Chat-v1.0 text-generation | 1,201 | 1.1B base model | — | Apache-2.0 | commercial ok |
| sammlapp/BirdNET_v2.4 | 1,192 | — | — | CC-BY-NC-SA-4.0 | no commercial |
| litert-community/Qwen3-ASR-0.6B automatic-speech-recognition | 1,157 | 938M base model | — | Apache-2.0 | commercial ok |
| Helsinki-NLP/opus-mt_tiny_eng-fra translation | 1,137 | 25M | — | Apache-2.0 | commercial ok |
| litert-community/Qwen3-TTS-12Hz-0.6B-Base text-to-speech | 1,109 | 915M base model | — | Apache-2.0 | commercial ok |
| Helsinki-NLP/opus-mt_tiny_fra-eng translation | 1,101 | 25M | — | Apache-2.0 | commercial ok |
| yolain/selfie_multiclass_256x256 | 1,032 | — | — | — | unknown |
| Helsinki-NLP/opus-mt_tiny_spa-eng translation | 1,028 | 25M | — | Apache-2.0 | commercial ok |
| soniqo/VoxCPM2-LiteRT-INT8 text-to-speech | 1,020 | 2.3B base model | 8-bit | Apache-2.0 | commercial ok |
| mailseth/coral | 1,016 | — | — | MIT | commercial ok |
Ordered by downloads, which on these 202,261 models is a real signal — 96% have a non-zero count. Parameter counts are resolved, not just read: the card first, then the base model's card, then the number in the name, then the GGUF file size, which is a per-parameter figure. That sizes 146 of the 312 models tflite can load; the source is printed under every count that did not come from the card itself. Quantisation is read the same way: the card's quant_bits or method where stated, otherwise the GGUF builds the repository actually ships, which is where a range like 2–16-bit and a build count come from.
The licence column resolved in full is on what you're allowed to ship. Everything here is queryable through the API.