TimesFM 2.0 500M
Model directorygoogle/timesfm-2.0-500m-pytorch500M params | Upstream checkpoint
TimesFM 2.0 500M is Google's larger open TimesFM checkpoint for zero-shot time-series forecasting, at 500M parameters. It is the higher-capacity, quality-tier member of the TimesFM line, focused primarily on univariate point forecasting with strong long-context behavior.
Architecturally it is a decoder-only patched transformer (the TimesFmModelForPrediction setup): history is split into patches and decoded into a forecast horizon. The base model targets point forecasts, with optional experimental quantile heads that Google notes are not calibrated after pretraining. Its pretraining builds on the TimesFM family corpus plus an added LOTSA subset listed by the official model card, on top of the broader Google Trends, Wikimedia Pageviews, and synthetic mixture used across the TimesFM line.
Where to run this model
Compute availability, hardware requirements, and pricing depend on the execution provider.
Get model weights, deployment instructions, and current provider listings.
- Retrocast · The Forecasting Company
managed service · Source checked 2026-10-06
Forecast through the managed API using timesfm-2. The provider controls the deployed variant and request limits.
- sktime TServe
self-hosted runtime · Family / provider variant · Source checked 2026-10-06
Run a model-serving container on your own hardware or compute account. Select the catalog checkpoint and matching runtime image.
Hugging Face provider mappings checked 2026-10-06. No Inference Provider mapping was listed for this checkpoint.
Model Classification
Family
TimesFM
Type
time series foundation model
Pretrained time-series model for zero-shot or few-shot forecasting workloads.
Resources
Training Data
TimesFM family pretraining corpus plus an added LOTSA subset listed by the official model card, building on the broader Google Trends, Wikimedia Pageviews, and synthetic mixture used by the TimesFM line.
Recommended For
- • Long-context zero-shot forecasting with strong open-model baselines
- • Workloads where point forecasting quality matters more than broad task coverage
Strengths
- • Large open checkpoints with long context windows
- • Efficient patched-transformer design with strong zero-shot behavior
Limitations
- • Primarily a forecasting family rather than a general multi-task TSFM
- • Quantile support is not the main identity of the family
Capabilities
Tags
Specifications
- Parameters
- 500M
- Architecture
- decoder-only patched transformer (TimesFmModelForPrediction)
- Context length
- n/a
- Max context
- n/a
- Minimum history
- n/a
- Recommended history
- n/a
- Input step
- n/a
- Required target series
- 1
- Temperature
- Ignored
- Top P
- Ignored
- Max output
- n/a
- License
- n/a