What does Nixtla do?
Nixtla is a time-series forecasting platform built on TimeGPT, its foundation model for forecasting and anomaly detection, alongside free open-source libraries such as StatsForecast. Enterprise pricing is quoted, not published. It suits data teams forecasting demand, sales or load; skip it if you need a no-code dashboard rather than an API and Python SDK.
- TimeGPT foundation model. A pretrained model for time series that produces forecasts and detects anomalies without training a model from scratch.
- Nixtla Engine. Described by Nixtla as a system that automatically builds the right forecasting pipeline for each use case.
- 200+ forecasting models. One interface for TimeGPT and the open-source Nixtlaverse models, with other foundation models such as Chronos and TimesFM named on the homepage.
- Anomaly detection. Historical and real-time (online) anomaly detection on time-series data.
- Fine-tuning. Set finetune_steps to adapt TimeGPT to your data; the model trains across all series in a dataset at once.
- Exogenous variables. Pass external drivers such as holidays, prices or categorical features through the X_df argument.
- Explainability and what-if. Driver analysis to explain forecasts and scenario planning for what-if questions.
- Self-hosting and integrations. Run it in your own infrastructure or Nixtla Cloud; connectors listed include Databricks, Snowflake, S3/GCS/ADLS, Pandas/Polars/Spark, Tableau, Power BI and Slack/Teams alerts. An MCP server is also shown.
Who is Nixtla for?
Nixtla is a good fit if you're a team lead, and you need help building and running AI apps.
How much does Nixtla cost?
Nixtla doesn't publish prices for its paid plans. It also has a free plan. Its pricing page lists a free trial. We checked this on Nov 5, 2026 on the official pricing page.
- Starts at
- Custom pricing
- Free plan
- Yes
- Free trial
- Free trial on request (length not stated)
Nixtla has no pricing page: the homepage lists a Free Forever tier (six open-source forecasting libraries) and Enterprise (premium models, fine-tuning, SLAs, dedicated support) with no price and a 'Request free trial' button. The TimeGPT API FAQ says paid plans are arranged through support once free API calls run out. Checked Nov 5, 2026.
Nixtla plans compared
| Plan | Billed monthly | Billed yearly | What you get |
|---|---|---|---|
| Free Forever (open source) | $0 | $0 | Six open-source libraries: StatsForecast, MLForecast, NeuralForecast, HierarchicalForecast, UtilsForecast and CoreForecast; you run them yourself |
| TimeGPT API (free calls) | Free to start | Not listed | Registration is open to everyone; once the free API calls are used up the API returns a 429 error and a paid plan is arranged with support. The number of free calls isn't stated |
| Enterprise | Custom pricing | Not listed | Premium models, fine-tuning on your data, SLAs and dedicated support; self-hosted (Docker, pip, Terraform) or Nixtla Cloud; free trial on request |
Checked on Nov 5, 2026 with a US-based fetch. Nixtla has no separate pricing page; the pricing options sit on the homepage, and Enterprise shows no price, only a 'Request free trial' button with no stated length. The TimeGPT docs FAQ says paid plans are arranged by contacting [email protected]. We list no dollar figures because none are published.
Checked Nov 5, 2026 on Nixtla's pricing page. Prices change often, so confirm before you buy.
Nixtla pros and cons
What stands out
- Free, open-source forecasting libraries you can use without a contract
- TimeGPT forecasts without training a model from scratch
- Supports fine-tuning and exogenous variables
- Self-hosted option keeps data in your own infrastructure
- Homepage lists SOC 2 Type II, SSO, RBAC and audit logs
- Python SDK with a short quickstart
Watch out for
- No published prices for paid API use or Enterprise
- Free API call allowance isn't stated
- Needs Python or API skills; not a no-code tool
- TimeGPT can't handle irregular timestamps or missing values
- Accuracy drops on long horizons, per its own FAQ
Based on Nixtla's own pages, not hands-on testing.
What are the alternatives to Nixtla?
We list 4 other AI Dev & Infrastructure tools: Netlify, Pinecone, Runpod, TheTop. Each one below shows its starting price and when you might pick it instead.
Compared in
Nixtla FAQ
How much does Nixtla cost?
As of Nov 5, 2026, Nixtla publishes no prices. Its open-source libraries are free, TimeGPT API registration starts free, and Enterprise is quoted by sales.
Is Nixtla free?
The open-source libraries (StatsForecast, MLForecast, NeuralForecast, HierarchicalForecast, UtilsForecast, CoreForecast) are free forever. The TimeGPT API also starts with free calls; the allowance isn't stated.
What is TimeGPT?
Nixtla's foundation model for time series. It forecasts and detects anomalies from your historical data through an API or Python SDK.
Does Nixtla have a free trial?
Enterprise has a 'Request free trial' button on the homepage. The trial length isn't stated.
What happens when the free TimeGPT calls run out?
According to the docs FAQ, the API returns a 429 error and you contact [email protected] for a paid plan.
Can I fine-tune TimeGPT?
Yes. Set finetune_steps; the model trains across all series in your dataset at once. Too many steps can overfit, the FAQ warns.
Can Nixtla run on my own servers?
Yes. A self-hosted version is available via Docker or a Python wheel, and the homepage also lists Terraform and a managed Nixtla Cloud.
Who is Nixtla best for?
Data and analytics teams forecasting demand, sales, energy or other time series in code. Teams wanting point-and-click forecasting in a BI tool may prefer a no-code option.
Sources and how we checked
- Official pricing page, checked Nov 5, 2026
- Official website
- Nixtla docs overview (read Nov 5, 2026), checked Nov 5, 2026
- TimeGPT FAQ: API key, free calls, fine-tuning, self-hosting (read Nov 5, 2026), checked Nov 5, 2026