t0-beta-q8_0-webgpu

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

theforecastingcompany/t0-beta, by The Forecasting Company

Q8_0 quantization for WebGPU by Idle Intelligence.

Apache-2.0

Matches or beats the official published t0-beta INT8 card: 0.20% worst-case mean drift against F32 versus 0.23% for the official card, and much tighter point drift (1.06% vs 9.39%). No browser measurement exists yet; only native Metal latency has been measured.

Type
model
Runs
browser, WASM, WebGPU
Size
256M params, 275.3 MB (Q8_0)
Demo (GitHub Pages)
https://idle-intelligence.github.io/t0-web/web/
Source
idle-intelligence/t0-web
HuggingFace
idle-intelligence/t0-beta-q8_0-webgpu
Status
maintained

← Home