Spike Model
SPIKEThe same architecture, sharpened on the tails — price and congestion spikes.
A GTFM derivative post-trained to model the extreme tails — the price and congestion spikes that matter most and are hardest to forecast.
How it works
Built on the Geotemporal Foundation Model backbone, then post-trained with a tail-weighted objective — quantile loss reweighted toward the extreme end of the distribution — over spike-labeled history, so model capacity concentrates on the upper quantiles instead of the easy middle. The output head is quantile-based rather than a single point forecast: it emits calibrated exceedance probabilities, the odds of clearing a given price or congestion threshold at each node and horizon, checked against realized outcomes to stay calibrated rather than just confident-looking.
Inputs
- All Geotemporal Foundation Model inputs
- Binding-constraint and shadow-price history, for the congestion tails
Outputs
- Spike probabilities and exceedance curves for price and congestion, by node and horizon
- Early-warning flags when tail risk builds
Three ways to consume it
product · inference · MCPIn the product
The Spike Model powers Ask the Grid directly — the live map, the radar, and the grid agent read it in place, so the same model you'd call is the one already reasoning over the grid on screen.
Developer inference service
A direct, typed inference API: request forecasts or embeddings by node and horizon and get structured results back. This is a plain inference endpoint — not MCP — for pipelines, notebooks, and services that want the model output without an agent in the loop.
See the developer surfaceMCP
The same model outputs, exposed as tools your own agent can call. Point a compatible client at the Ask the Grid MCP server and the model becomes a tool in your agent's own loop.
Build with MCPIntended use
- Spike and scarcity early warning
- Congestion-risk assessment on specific paths and nodes
- Hedging and risk desks pricing the tails
Limitations
- Forthcoming — not generally available; capabilities described are forward-looking
- Outputs are probabilistic exceedance estimates, not guaranteed events
- Inherits the Geotemporal Foundation Model's per-ISO coverage rollout