H2O.ai’s TimeH2O-v1 tops time-series benchmarks with 91.9% win rate
It achieved a 91.9% win rate and a 50.8% lower forecast error than a seasonal-naive baseline across 100 forecasting tasks. The model handles univariate series and covariates, covering finance, healthcare, supply chain, and energy—without fine-tuning on fev-bench data. TimeH2O-v1 has 108 million parameters and is available on Hugging Face under a non-commercial license for internal testing.…
Key points
- TimeH2O-v1 leads fev-bench with 91.9% win rate, 50.8% lower error than baseline across 100 tasks
- Model handles univariate series and covariates (e.g., price, weather) with 108M parameters
- Available on Hugging Face under non-commercial license; commercial access requires direct contact
The company positions TimeH2O-v1 as a zero-shot forecasting tool that organizations can fine-tune for domain-specific use. H2O.ai emphasizes its ‘Sovereign AI Platform,’ which supports production deployment across cloud and air-gapped environments. The model’s performance aligns with its mission to democratize AI forecasting for regulated sectors like finance and healthcare, where accuracy and governance are critical.
Model pages: Timestral-209M → · TimesFM-3 →
The story so far
2 episodes →- H2O.ai’s TimeH2O-v1 tops time-series benchmarks with 91.9% win ratethis story
H2O.ai's TimeH2O Beats Mistral, Google, and Amazon in Time-Series Forecasting
aol.com · 26 October 2026
Loading the full article…
This text was published by aol.com and written by AOL. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- IEEE launches AI Research Suite for Xplore Digital Library · 1 src
- OpenAI releases ChatGPT work usage data for 165 activities · 1 src
- OpenAI details layered protections in AI research environments · 1 src
- Foundation models match but sometimes lag task-specific models in prostate pathology · 1 src
- Researchers release ProBEiT, a 400M-parameter protein model · 1 src
Comments
via GitHub Discussions