IT'S ABOUT TIME.
Introducing Time Series Language Models (TSLMs)
Every heartbeat, price tick, and sensor stream creates time series data. In contrast to text, image, and audio, frontier AI models did not understand time series data, until now.
OUR MISSION
Time series is the language of the physical economy.
As a frontier AI company, we are the first to build foundational Time Series Language Models that input and output text and time series: AI that understands temporal sensor data natively, incorporates context, turns signals into action, and enables prompt-based interaction with and generation of time series data.
While Time Series Foundation Models (TSFMs) exist, they are not language models. TSFMs perceive raw signals but cannot explain them. In contrast, LLMs can reason in natural language but cannot perceive time series. A TSLM is one model that does both. A new modality for AI. TSLMs enable direct reasoning, explanation, and forecasting over temporal data in natural language, with order-of-magnitude gains in temporal reasoning on smaller, faster backbones compared to existing frontier models.
Our first model, OpenTSLM, is accepted at ICML 2026 and scores 99.71 on the TSQA reasoning benchmark where GPT-4o reaches 59.24. It is released open source under the MIT license.
| TSFMsTime Series Foundation Models | LLMs / VLMsLLMs and VLMs | OpenTSLMTime Series Language Models | |
|---|---|---|---|
| Perceives raw time series | Yes | No | Yes |
| Learns multi-variate dependencies | Yes | No | Yes |
| Reasons in natural language | No | Yes | Yes |
| One model across domains & tasks | No | Yes | Yes |
| Human steering: explains itself, takes commands | No | Yes | Yes |
DETECT
Catch what dashboards miss: subtle drifts, correlations, and anomalies surfaced before they become failures.
EXPLAIN
Every finding in plain language: what happened, when, why, and how confident the model is.
PREDICT
See what comes next: how the signal evolves, when limits are reached, and what changes under intervention.
ACT
Recommended interventions with the expected outcome of each, so the expert decides with the full picture.
VISION
What we're building
Our vision is to build the layer that connects continuous real-world signals to intelligent decisions and autonomous agents. A universal TSLM will power proactive healthcare, adaptive robotics, resilient infrastructure, and new forms of human-AI collaboration: the temporal interface for AI.
CUSTOMER NEEDS
Pain points
Operational data
Deployment constraints
Horizontal
TSLM Platform
INGESTION → TRAINING →
EVALUATION → DEPLOYMENT / API
REUSABLE APPLICATIONS
Predictive maintenance
Anomaly detection
Forecasting
Optimization
TEAM
Who we are
We are a team of scientists, engineers, and builders with experience across leading research institutions, frontier technology companies, and global industrial organizations. We are the original authors of OpenTSLM, the first open framework for language models that reason over time series data.
COLLABORATIONS
Partner with us
We are looking to collaborate with organizations interested in developing and validating high-value applications across healthcare, finance, industry, energy, transportation, and other data-rich domains. Apply to join our selective enterprise pilot program, tailored to your workflows and target outcomes.