TimesFM-3
TimesFM-3 is a released general LLM work model with open-weight; evaluate it while provider pricing coverage matures.
Use it for
- Teams evaluating general LLM work
Do not use it for
- Cost-sensitive launches that need sourced token pricing
- Vision or document-understanding workloads
- Strict JSON or tool-calling flows
- Family
- TimesFM
- Released
- 2026-08-31
- Parameters
- 330M
- Architecture
- Decoder Only
- Specialization
- forecasting
- Openness
- Open weights
- License
- NoncommercialCommercial use: non-commercial
- Weights
- Available
- Code
- Unknown
- Training
- Pretrained
No tracked provider token pricing is available yet.
About
TimesFM-3 (TimesFM 3.0) is Google Research's zero-shot time-series foundation model for multivariate and univariate forecasting, released August 31, 2026. First-party blog: 330 million parameters, pretrained on more than 1 trillion real-world and synthetic time points; decoder-only transformer with 32-step patches, alternating causal temporal attention and full variate attention, and single-pass Contiguous Patch Masking. Natively supports multiple targets, past covariates, and past-future (dynamic) covariates, with point forecasts plus 9 quantiles (10th–90th percentile). Official PyTorch checkpoint google/timesfm-3.0-pytorch on Hugging Face. GitHub code is Apache-2.0; TimesFM 3.0 pretrained weights are TimesFM Non-Commercial License v1.0 (non-commercial, non-production).
Top use-case fit
No primary decision-task fit is mapped for this model yet.
Provider price ladder
No tracked provider token pricing is available for this model yet.
Capabilities
No model capability flags are currently sourced.
Benchmark peer barsfor Coding
No task-mapped benchmark peers are available for this model yet.
Migration checks
No linked migration route is available for this model yet.
API versions
timesfm-3No tracked provider token pricing is available yet.