Collection

Every model, dataset, and repo that came out of the lab.

Speech-to-text

stt-web

Transcribes speech to text live in the browser, in English and French.

70ms/frame, RTF 0.91-0.95x

1 model browser

LLM and tool calling

agent-web

A tool-calling agent loop for Qwen2.5-Instruct models, running on the lean inference engine.

browser

LLM and tool calling

llm-web

lean, a small LLM inference engine in Rust and WGSL that runs the same code natively and in the browser.

browser

Text-to-speech

tts-web

Converts text to speech in the browser with two lightweight voice models; Pocket TTS speaks six languages.

2.28x realtime, WASM/Chrome

3 models browser

Speech-to-speech

sts-web

A full-duplex speech-to-speech assistant running entirely client-side, still early and rough.

24 of 32 layers, Q4_K

2 models browser

Audio codec and VAD

mimi-rs

A Rust implementation of an audio codec that tokenizes and reconstructs speech.

96.2M params

browser

Audio codec and VAD

vad-rs

Detects the start and stop of speech in an audio stream using a small neural model.

1 model browser

Forecasting

t0-web

A from-scratch port of a probabilistic time-series forecasting transformer to the browser.

30-34ms per forecast, Chrome

4 models browser

Cellular automata

llm-life

Uses a small language model's logits as the update rule for Conway's Game of Life.

16384²: GPU ~790× the CPU

2 models browser

Text classification

hive-ml-classifier

A small BERT-mini classifier that routes chat messages into intent categories.

94.4% accuracy, 94.3% F1

1 model browser

Terrain and visualization

ridgeline

A 3D globe visualizer that renders planetary terrain as stacked ridgeline plots.

11 bodies rendered

1 dataset browser

Weather

weather-web

A weather estimate for any point on Earth from the nearest METAR-reporting stations.

7,534 stations worldwide

1 dataset browser

Sensor and privacy research

sensors-web

A page that probes which device sensors a mobile browser exposes without a permission prompt.

browser

Benchmarking

nordcoop-bench

A reproducible benchmark comparing local language models against hosted references on constrained hardware.

33.2 t/s decode, Qwen3-Next-80B-A3B

native

Distributed inference

Hive

A swarm of small language models running in browser tabs, with a hub that routes questions to them.

browser

Speech-to-text

stt-1b-en_fr-q4_0-webgpu

Q4-quantized weights for a ~1B-parameter streaming speech-to-text transformer, packaged for the browser.

from stt-web

browser

Text-to-speech

kitten-tts-nano-safetensors

Voice weights for a small text-to-speech model, converted to a Rust-friendly format.

from tts-web

browser

Text-to-speech

pocket-tts-gguf

The GGUF Q8_0 weights, one per language, that tts-web's Pocket TTS engine loads in the browser.

from tts-web

browser

Text-to-speech

pocket-tts-int8

An INT8-quantized version of a lightweight autoregressive text-to-speech model.

from tts-web

browser

Speech-to-speech

personaplex-24L-q4_k-webgpu

The pruned, QLoRA-recovered, Q4_K-quantized PersonaPlex-7B that sts-web's live demo actually loads.

from sts-web

WebGPU

Speech-to-speech

personaplex-7b-v1-q4_k-webgpu

An unofficial quantization of a large speech-to-speech model for browser deployment.

from sts-web

WebGPU

Audio codec and VAD

silero-vad-v5-safetensors

Silero VAD v5 weights, converted from ONNX to safetensors for candle inference.

from vad-rs

native

Forecasting

t0-alpha-q4_0-webgpu

Q4_0-quantized weights for t0-alpha, packaged for browser WebGPU forecasting.

from t0-web

browser

Forecasting

t0-alpha-q8_0-webgpu

Q8_0-quantized weights for t0-alpha, packaged for browser WebGPU forecasting.

from t0-web

browser

Forecasting

t0-beta-q4_0-webgpu

Q4_0-quantized weights for t0-beta, packaged for browser WebGPU forecasting.

from t0-web

browser

Forecasting

t0-beta-q8_0-webgpu

Q8_0-quantized weights for t0-beta, packaged for browser WebGPU forecasting.

from t0-web

browser

Cellular automata

llm-of-life-lora

LoRA adapters that turn Qwen2.5-0.5B-Instruct into a Game of Life update rule.

from llm-life

browser

Cellular automata

stencil-life

The smallest from-scratch networks trained to reproduce Conway's Game of Life exactly.

from llm-life

browser

Text classification

inference-at-home

The trained weights behind hive-ml-classifier's swarm-intent BERT.

from hive-ml-classifier

browser

Go back home.