IBM Granite · hosted on Ephemeris
PatchTST-FM r2 API
IBM Granite's patch-transformer foundation model. The strongest single member at GIFT-Eval's long horizons in our run.
Explicit mode
"model": "patchtst-fm-r2"API namepatchtst-fm-r2
PublisherIBM Granite
Weightsibm-granite/granite-timeseries-patchtst-fm-r2
Size385M
LicenceOpenMDW-1.0
MultivariateYes
CovariatesNo
Context read8,192 most recent points
01
REST
Call PatchTST-FM r2 directly
cURLUTF-8
curl --request POST \
--url https://ephemeris.cascade.industries/api/v1/forecast \
--header "Authorization: Bearer pc_live_your_key" \
--header "Content-Type: application/json" \
--data '{
"mode": "explicit",
"model": "patchtst-fm-r2",
"series": [{ "values": [112, 118, 132, 129, 121, 135, 148, 148], "freq": "M" }],
"horizon": 12,
"quantiles": [0.1, 0.5, 0.9]
}'02
Agents
From Claude, Cursor or any MCP client
Add the Ephemeris MCP server, then ask your agent to forecast with patchtst-fm-r2. The forecast tool takes mode: "explicit", model: "patchtst-fm-r2".
03
Accuracy
How it scores
On GIFT-Eval (97 configurations, official harness), PatchTST-FM r2 alone scores CRPS 0.4869 and MASE 0.7150 as a ratio to seasonal naive (lower is better). The Ephemeris ensemble, which includes it, scores 0.4662 and 0.6841. Full benchmark results.
No GPU to run