TSFM.ai developer documentation.
Multiple pages, one contract. API, MCP, and CLI are aligned on the same schema so teams can move from manual calls to production automation with zero drift.
Learn
Choosing the right model
TSFM.ai offers 17 models across different architectures, sizes, and specializations. This guide helps you pick the right one based on your latency, cost, accuracy, and task requirements.
If you are not sure where to start
Start with Chronos-Bolt Base for most workloads. It offers a strong balance of accuracy, speed (130ms), and cost ($0.07/1M tokens) with probabilistic output support. From there, move to TimesFM 2.5 if you need longer context, Moirai 2.0 for multivariate, or Granite TTM if you need to minimize cost.
Decision factors
Key dimensions to consider when selecting a model.
| Factor | Low | Mid | High |
|---|---|---|---|
| Latency requirement | < 150ms | 150-400ms | > 400ms |
| Budget per 1M tokens | < $0.10 | $0.10-0.25 | > $0.25 |
| Context needed | < 5K tokens | 5K-12K | > 12K |
| Task complexity | Point forecast | Probabilistic | Multi-task |
Recommendations by scenario
Lowest latency
You need sub-150ms responses for real-time dashboards, alerting, or streaming applications.
Recommended
These models use direct prediction or lightweight architectures that minimize inference time.
Lowest cost
You are processing millions of series in batch and need to minimize per-request cost.
Recommended
Smaller parameter counts mean lower GPU utilization per request. Combined with high rate limits, these are ideal for batch workloads.
Best forecast quality
Accuracy is the primary concern and you can tolerate higher latency and cost.
Recommended
Larger models with longer context windows capture more complex patterns. Moirai Large excels at multivariate, TimesFM at long context, TiRex at covariate-heavy data.
Multivariate series
You have multiple correlated variables that should be modeled jointly.
Recommended
Moirai's Any-Variate Attention captures cross-variate dependencies natively. Time-MoE's expert routing handles diverse multivariate domains.
Anomaly detection
You need to detect unusual patterns in monitoring, sensor, or transactional data.
Recommended
MOMENT is purpose-built for multi-task including anomaly detection. Toto is specialized for infrastructure metrics from Datadog's telemetry.
Limited history
You have fewer than 50 historical observations (new products, new sensors).
Recommended
These models have strong zero-shot transfer from pre-training and produce reasonable forecasts even with minimal context.
Covariates and external signals
Your forecasts depend on external factors like promotions, holidays, or weather.
Recommended
TiRex has native covariate encoding. Chronos-2 added covariate support in v2. Both accept past and future covariates in the request payload.
Full model comparison
All 17 models sorted by latency. Click any model name to see full details.
| Model | Params | Latency | Input cost | Context | Best for |
|---|---|---|---|---|---|
| Chronos-Bolt Small | 48M | 88ms | $0.04 | 6K | Real-time and batch applications at lowest cost |
| Granite TTM 1M | ~1M | 95ms | $0.03 | 4K | Ultra-low-cost batch forecasting and edge deployment |
| PatchTST Large | ~40M | 115ms | $0.06 | 4K | Stable baseline for benchmarking and comparison |
| Chronos-Bolt Base | 205M | 130ms | $0.07 | 8K | Fast inference with strong accuracy |
| Sundial | 128M | 140ms | $0.08 | 8K | Calibrated uncertainty estimates (diffusion-based) |
| Moirai 2.0 Small | 14M | 210ms | $0.09 | 8K | Low-cost multivariate forecasting |
| Chronos-2 | 120M | 240ms | $0.16 | 8K | General-purpose probabilistic forecasting with covariate support |
| Time-MoE 200M | 200M/2.4B | 240ms | $0.11 | 12K | Cross-domain transfer with expert specialization |
| Toto | 151M | 250ms | $0.13 | 8K | Observability and infrastructure telemetry signals |
| TiRex Large | ~300M | 260ms | $0.15 | 16K | Covariate-heavy workloads (holidays, promotions) |
| Lag-Llama | ~10M | 280ms | $0.14 | 8K | Full predictive distributions with uncertainty |
| TimesFM 2.5 | 200M | 290ms | $0.18 | 16K | Long-context forecasting with large historical windows |
| Moirai 2.0 Base | 91M | 330ms | $0.17 | 12K | Balanced multivariate quality and cost |
| MOMENT Large | 385M | 410ms | $0.20 | 12K | Multi-task: forecast + anomaly + classify + impute |
| Moirai 2.0 Large | 311M | 520ms | $0.29 | 16K | Maximum multivariate forecast quality |
| Time-LLM 7B | 7B | 650ms | $0.32 | 8K | Research and LLM-based reasoning over series |
Browse models
See all models with live status, pricing, and detailed specifications.
Model catalog
Quickstart
Make your first API call and see forecasts in under 5 minutes.
Get started
Playground
Test models interactively with your own data before committing to an integration.
Try it now