RealFeedApp Four Timestral versions from 5M to 209M parameters run efficiently on CPUs
The new Timestral model ranks first on three major benchmarks using only two-thirds of the parameters of previous leaders and runs efficiently on CPUs.
- Field
- training language models
- What they did
- Mistral released the Timestral model, which makes time-series forecasts using only 209 million parameters, outperforming larger models on key benchmarks.
- Why it matters
- This enables accurate predictions in various fields while ensuring the model runs efficiently on standard CPUs.
#timestral#mistral#time-series#forecasting#foundation model
