Audience
Data scientists, researchers, and developers wanting to forecast multiple related time series and incorporate historical and known future signals
About TimesFM-3
TimesFM-3 is a state-of-the-art time series foundation model designed for highly accurate multivariate forecasting in a single forward pass. The 330 million parameter model is pre-trained on a real-world and synthetic time-series corpus comprising more than 1 trillion time points, building on the efficiency and zero-shot generalization of earlier TimesFM models. It can jointly predict multiple coevolving time series and capture dependencies that improve accuracy without task-specific fine-tuning. The model supports multiple targets with point and quantile forecasts, past covariates that are known only historically, and past-future dynamic covariates such as planned promotions, holidays, or weather forecasts. TimesFM-3 uses a decoder-only transformer architecture, processes contiguous data in patches of 32 time steps, and applies alternating causal temporal attention and full variate attention to combine patterns across time and related series.
Other Popular Alternatives & Related Software
Kimi K3
Kimi K3 is Moonshot AI’s most capable model, built for frontier intelligence scenarios such as software engineering, knowledge work, deep reasoning, and multimodal understanding. The model has 2.8 trillion parameters and uses Kimi Delta Attention, a hybrid linear attention mechanism, along with Attention Residuals for long-context performance. Kimi K3 supports a 1 million token context window, making it useful for analyzing large codebases, long documents, complex knowledge bases, and multi-step workflows. It includes native visual understanding for images and videos, with support for structured message formats, base64 image input, uploaded video files, and multimodal reasoning. Developers can use Kimi K3 through an OpenAI-compatible API with support for streaming, structured JSON output, partial mode, custom tools, dynamic tool loading, and automatic context caching.
Learn more about Kimi K3
Nixtla
Nixtla is a platform for time-series forecasting and anomaly detection built around its flagship model TimeGPT, described as the first generative AI foundation model for time-series data. It was trained on over 100 billion data points spanning domains such as retail, energy, finance, IoT, healthcare, weather, web traffic, and more, allowing it to make accurate zero-shot predictions across a wide variety of use cases. With just a few lines of code (e.g., via their Python SDK), users can supply historical data and immediately generate forecasts or detect anomalies, even for irregular or sparse time series, and without needing to build or train models from scratch. TimeGPT supports advanced features like handling exogenous variables (e.g., events, prices), forecasting multiple time-series at once, custom loss functions, cross-validation, prediction intervals, and model fine-tuning on bespoke datasets.
Learn more about Nixtla
GLM-5.3-Flash
GLM-5.3-Flash is Z.ai’s natively multimodal model in the GLM-5 series (previously previewed as Ox Alpha), designed to deliver strong coding, agentic, visual, and knowledge-work performance at relatively low inference cost. It uses 320 billion total parameters with 18 billion active parameters, along with a hybrid architecture that combines sparse and linear attention to reduce the cost of long-context processing. The model supports context lengths of up to one million tokens and was trained on a 30-trillion-token multimodal corpus. GLM-5.3-Flash can reason across text, images, documents, interfaces, dashboards, and other visual information while using that feedback to refine its own outputs. Z.ai reports substantial gains over GLM-5.2 on coding and agentic benchmarks, including DeepSWE and AutomationBench, while approaching higher-cost frontier models on several evaluations.
Learn more about GLM-5.3-Flash
Timestral
Timestral is a family of time-series foundation models and a Python inference package for generating multivariate probabilistic forecasts from historical sequences. The four model sizes (5M, 28M, 123M and 209M parameters) support past-only, known-future and categorical covariates, return quantile forecasts, and accept context windows up to 32,768 time steps. Developers can load pretrained weights from the Hugging Face Hub or from local files and run inference on CPU or GPU, then incorporate forecasts into monitoring, capacity-planning, energy, supply-chain, financial-risk, or maintenance workflows. Timestral supports zero-shot forecasting and can be adapted to domain data through Mistral Forge. The package is intended for data scientists, machine-learning engineers, and teams building forecasting systems; repository code is Apache 2.0, while model weights use a separate Modified MIT license.
Learn more about Timestral
Integrations
No integrations listed.
Company Information
Google
Founded: 1998
United States
research.google/blog/timesfm-3-a-zero-shot-foundation-model-for-multivariate-forecasting/
Videos and Screen Captures
Other Useful Business Software
Our Free Plans just got better! | Auth0
You asked, we delivered! Auth0 is excited to expand our Free and Paid plans to include more options so you can focus on building, deploying, and scaling applications without having to worry about your security. Auth0 now, thank yourself later.
Product Details
Platforms Supported
Cloud
Training
Documentation
Videos
Support
Online
TimesFM-3 Frequently Asked Questions
Q: What kinds of users and organization types does TimesFM-3 work with?
TimesFM-3 works with these users and organization types: Mid Size Business, Small Business, Enterprise, Freelance, Nonprofit, Government, and Startup.
Q: What languages does TimesFM-3 support in their product?
TimesFM-3 supports these languages: English.
Q: What kind of support options does TimesFM-3 offer?
TimesFM-3 offers support via online.
Q: What type of training does TimesFM-3 provide?
TimesFM-3 provides training in the form of documentation and videos.
