Model directory reference
Read model identities, research facts, and execution options without confusing them with a TSFM deployment.
Identity and upstream source
Each model page represents a checkpoint or family member, with an identifier, architecture, parameter count when available, license, research links, and upstream Hugging Face repository. Model authors are not necessarily execution providers. Preserve the exact model revision when you evaluate or deploy it.
Capabilities and runtime support
Upstream task support describes what the original implementation can do. An export or endpoint may expose a smaller contract. Browser support is checkpoint-specific and requires numerical validation; a family name or an ONNX tag is insufficient. An encoder-only export cannot provide a forecast simply because the original model can forecast.
Find an execution option
Use the browser action where enabled, or follow Hugging Face links for current provider and deployment options. A model on the Hub is not automatically available from an Inference Provider. TSFM does not publish its former hosted status, price, latency, GPU region, or uptime as model specifications.