pretickGoogle TimesFM 2.5 (200M)
A general forecasting perspective built around numerical sequences, rather than financial news or company fundamentals. We do not infer financial specialization from the scale of its pretraining.
Training background
The original TimesFM research describes pretraining on a broad mixture of real and synthetic time series. That paper covers an earlier generation: it is not a complete disclosure of the dataset mix used for the 2.5 checkpoint.
Available through Pretick
Pretick uses the native PyTorch 200M checkpoint. Each target is forecast separately, with up to 2,048 input rows and 1,024 future points. This integration does not use past covariate columns; such requests exclude TimesFM from consensus.
{
"model": "google/timesfm-2.5",
"data": [
102,
108,
111,
104,
99,
106
]
}Use this request at POST https://pretick.ai/v1/timeseries/forecast, or select the same model through MCP. Omit model for consensus. Each successful request costs $0.02.
Using this model for price forecasting
Reading the evidence
The financial-return study in our reference library includes TimesFM 2.5. Its results are specific to the tested equities, periods and evaluation setup, not a ranking for every market.
Sources and version
Hosted checkpoint: google/timesfm-2.5-200m-pytorch
Revision: 1d952420fba87f3c6dee4f240de0f1a0fbc790e3
License/access: Apache 2.0.
Responses identify the exact model version. These pages describe Pretick's integration; the original model may support additional features.