The inference layer for time series foundation models
TSFM.ai is a unified API gateway that gives developers instant access to the world's best open-weight time series foundation models. Upload a series, pick a model, get a forecast — we handle the GPUs, the tokenization, and the model routing so you can focus on your application.
Why TSFM.ai exists
Time series foundation models are one of the most exciting developments in applied ML — but running them in production is still surprisingly hard. Each model has its own tokenizer, input format, quantile convention, and GPU requirements.
We built TSFM.ai so that teams can evaluate and deploy any TSFM through a single, consistent API — the same way developers use OpenRouter or Together for LLMs. One API key, one schema, sixteen models.
16+
Models hosted
<300ms
Avg latency
99.9%
API uptime
50+
Research papers tracked
What we believe
Model-agnostic
We host the best open-weight time series foundation models from Amazon, Google, Salesforce, IBM, CMU, and more — so you can compare and switch without vendor lock-in.
Transparent pricing
Per-token pay-as-you-go pricing with no hidden fees, no seat licenses, and no minimum commitments. Every model's cost is published on the catalog page.
Research-first
We track the frontier of time series foundation model research and bring new architectures to production as soon as they prove out on public benchmarks.
Developer experience
One unified API across all models. OpenAI-compatible endpoints, first-party MCP server, CLI tooling, and SDKs — so you can go from zero to forecast in minutes.
How it works
- 1
Get an API key
Sign up for free. Your first 1,000 calls per month cost nothing.
- 2
Send a time series
POST your historical observations to /v1/forecast with any model from the catalog.
- 3
Get predictions
Receive point forecasts, quantile intervals, and metadata — all in a single JSON response.
Ready to start forecasting?
Create a free account, grab an API key, and make your first forecast in under a minute.