ModelsPublished All news from that day
H2O.ai's h2ofm package ships TimeH2O forecasting models of 108.8M and 5.8M parameters
The package is Apache-2.0, the weights are licensed separately, and the page gives no benchmark results.
Check our sources · 6 facts from 1 sourceH2O.ai's h2ofm Python package, version 0.43.1 released on 27 October 2026, runs standalone inference for the company's TimeH2O time-series forecasting models. The package page lists two models: h2oai/timeh2o-v1 with 108.8 million parameters and h2oai/timeh2o-v1-tiny with 5.8 million.
The package is licensed Apache-2.0, and the page says model weights are downloaded and licensed separately. It needs Python 3.12 or later and runs on CPU or CUDA. The page states no benchmark results.
Underlined sentences link to the facts we checked. Dotted words open a short explainer.
Fact 4 of 6 · from H2O.ai (PyPI), 27 Oct 2026
Apache-2.0; see the `LICENSE` file in the distribution. Upstream attribution: `THIRD_PARTY.md`, shipped alongside it. Model weights are downloaded and licensed separately.
Your reaction
We count reactions per story and day, never who reacted. The counts help us choose what goes in the monthly issue. If you are signed in, your own page shows yours too.
We count reactions per story and day, never who reacted. They help choose what goes in the monthly issue.
Check our sources
Every sentence above is checked against this source.
1 h2ofm: standalone inference for H2O.ai foundation models (TimeH2O)
Open the source archived copy-
The PyPI page for h2ofm lists version 0.43.1 as released on 27 October 2026, after 0.43.0 on 22 October and 0.42.0 on 13 October 2026.
cite0.43.1 uploaded 2026-10-27T13:46:12Z; 0.43.0 2026-10-22T21:20:06Z; 0.42.0 2026-10-13T12:39:33Z (PyPI JSON API release upload times)
-
The h2ofm page describes TimeH2O as time-series forecasting, supporting univariate and multi-target series and covariates, and returning point (median), quantile and mean forecasts.
citea 2D array batches univariate series; a 3D array batches multi-target series.
-
The h2ofm page lists two models: h2oai/timeh2o-v1 with 108.8M parameters and h2oai/timeh2o-v1-tiny with 5.8M parameters.
cite| `h2oai/timeh2o-v1` | 108.8M |
-
The h2ofm package is licensed Apache-2.0; the page says: "Model weights are downloaded and licensed separately."
citeApache-2.0; see the `LICENSE` file in the distribution. Upstream attribution: `THIRD_PARTY.md`, shipped alongside it. Model weights are downloaded and licensed separately.
-
The h2ofm page lists "Python 3.12+; CPU and CUDA" as requirements.
citePython 3.12+; CPU and CUDA.
-
The h2ofm page states no benchmark or accuracy results for TimeH2O.
citeStandalone inference for H2O.ai foundation models. Python 3.12+; CPU and CUDA.
Every sentence above is checked against this source. Each fact has its own address you can cite.
Source 1 · H2O.ai (PyPI) · 27 Oct 2026
h2ofm: standalone inference for H2O.ai foundation models (TimeH2O)
The PyPI page for h2ofm lists version 0.43.1 as released on 27 October 2026, after 0.43.0 on 22 October and 0.42.0 on 13 October 2026.
0.43.1 uploaded 2026-10-27T13:46:12Z; 0.43.0 2026-10-22T21:20:06Z; 0.42.0 2026-10-13T12:39:33Z (PyPI JSON API release upload times)
The h2ofm page describes TimeH2O as time-series forecasting, supporting univariate and multi-target series and covariates, and returning point (median), quantile and mean forecasts.
a 2D array batches univariate series; a 3D array batches multi-target series.
The h2ofm page lists two models: h2oai/timeh2o-v1 with 108.8M parameters and h2oai/timeh2o-v1-tiny with 5.8M parameters.
| `h2oai/timeh2o-v1` | 108.8M |
Show all 6 factsShow fewer facts
The h2ofm package is licensed Apache-2.0; the page says: "Model weights are downloaded and licensed separately."
Apache-2.0; see the `LICENSE` file in the distribution. Upstream attribution: `THIRD_PARTY.md`, shipped alongside it. Model weights are downloaded and licensed separately.
The h2ofm page states no benchmark or accuracy results for TimeH2O.
Standalone inference for H2O.ai foundation models. Python 3.12+; CPU and CUDA.
Topics
The morning email
On the mornings we publish: the three top stories and up to four short ones. Free.
We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep