perspectiva aiqbenchMachine learning
analiza aiqbench
Mistral's new Timestral family introduces a specialized foundation model for time series forecasting that operates efficiently on CPUs, challenging established benchmarks like Google's TimesFM. With parameter counts ranging from 5M to 209M, this release expands the practical toolkit available for developers working on predictive analytics without requiring high-end GPU clusters. This development signals growing maturity in domain-specific AI models beyond general-purpose LLMs.
1
Sursă care acoperăMarkTechPost
Mistral AI Releases Timestral: A 209M Forecasting Model Family That Outperforms Google’s TimesFM-3
MarkTechPost1 min de cititArticolul și perspectiva sunt în engleză.
Mistral released Timestral, a time series foundation model family from 5M to 209M parameters. The 209M model leads GIFT-Eval, TIME and fev-bench among zero-shot models and runs on CPU. The post Mistral AI Releases Timestral: A 209M Forecasting Model Family That Outperforms Google’s TimesFM-3 appeared first on MarkTechPost.
Citește articolul complet pe MarkTechPostSubiecte conexe
Machine learning
A Benchmark for Electrical Load Forecasting Across Grid Levels: Time-Series Transformers Outperform Established Methods
arXiv · 20 iul. 2026
Machine learning
Two Heads Are Better Than One: Aggregating Weaker LLMs for Better Forecasts
arXiv · 30 sept. 2026
Machine learning
Global probabilistic forecasting of air pollution
arXiv · 20 oct. 2026
Cercetare AI
Competence-Gated Pooling of Language Models and Priors for Event Forecasting
arXiv · 14 sept. 2026