Raspberry Pi 5
The go-to single-board computer for robots and voice devices that need small speech, detection and depth models without a GPU.
Models estimated to fit
15 indexed models with an estimated tier of Single-board computer or lower.
Estimated from parameter counts, not benchmarks. Measured a model on this board? Share the numbers in c/edge-ai.
Vision 3
- ustc-community/dfine-small-coco
Small real-time DETR-style detector for the 80 COCO classes that refines box edges as distributions for tighter localization.
- depth-anything/Depth-Anything-V2-Small-hf
Lightweight monocular model that predicts a relative depth map from a single RGB frame.
- facebook/dinov3-vits16-pretrain-lvd1689m
Smallest ViT from Meta's DINOv3 family, a self-supervised backbone whose dense patch features transfer to many vision tasks without fine-tuning.
Multimodal 1
- HuggingFaceTB/SmolVLM2-500M-Video-Instruct
Half-billion-parameter vision-language model from Hugging Face that answers questions about images and short video clips.
Audio 9
- moonshine-ai/moonshine-streaming-small
The small size of the Moonshine Streaming family, an English-only encoder-decoder recognizer that uses a sliding-window encoder so it can transcribe live audio with low latency on edge hardware.
- ggerganov/whisper.cpp
GGML-format conversions of the OpenAI Whisper checkpoints, from tiny to large, for the whisper.cpp C/C++ inference engine.
Works with GotchiGotchi Local can download this as an offline speech fallback on your computer. - hexgrad/Kokoro-82M
An 82M-parameter text-to-speech model built on StyleTTS 2 components, shipped with a set of preset voices across several languages.
- rhasspy/piper-voices
The official collection of ONNX voice models for the Piper offline TTS engine, covering many languages at several size and quality levels.
Works with GotchiGotchi Local can download this as an offline speech fallback on your computer. - kyutai/pocket-tts
A 100M-parameter CPU-oriented TTS model from Kyutai with voice cloning from a reference clip and support for English plus five other European languages, designed to run faster than real time using two CPU cores.
- onnx-community/silero-vad
ONNX exports of the Silero voice activity detector in fp32, fp16 and 8-bit forms (the int8 file is about 0.6 MB) that mark which short audio frames contain speech.
- pipecat-ai/smart-turn-v3
An 8M-parameter multilingual turn-detection model, a Whisper Tiny encoder with a linear head, that predicts from raw audio whether the speaker has finished; the int8 CPU ONNX build is about 8.7 MB.
- davidscripka/openwakeword
Pretrained openWakeWord models, small classifiers on top of a shared speech-embedding front end, that detect phrases such as 'hey jarvis' in a continuous audio stream.
- pyannote/speaker-diarization-community-1
The open speaker-diarization pipeline released with pyannote.audio 4.0, combining a segmentation model, speaker embeddings and PLDA-based clustering to label who spoke when, with better speaker counting than the 3.1 pipeline.
Language 2
- Qwen/Qwen3-0.6B
Smallest dense model in the Qwen3 family, with optional thinking mode and function-calling support.
- google/gemma-3-270m-it
Tiny instruction-tuned Gemma 3 model meant to be fine-tuned for narrow jobs such as classification, extraction or command routing.