Timestral 123M, int8 ONNX

An ONNX export of mistralai/Timestral-123M, quantized to int8 (dynamic, per-channel; embedding and output head kept in fp32), for running with ONNX Runtime on CPU.

File: timestral-123m-int8.onnx, 149,151,243 bytes
SHA-256: 8badce0571b688e4b5773a519bd9aad472151cf03982aa9fbad862f7666c66d4
One graph for horizons up to 32 steps, without covariates; normalization, mirror averaging, non-negative clip and quantile cap are inside the graph.

This is a derivative of Mistral AI's weights and is distributed under their Modified MIT License (see LICENSE, attribution notice "2026 - Mistral AI"), including its revenue condition.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

Time Series Forecasting

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Cosmicosrl/Timestral-123M-ONNX

Base model

mistralai/Timestral-123M

Quantized

(6)

this model