本文为原文前 6,000 字符的节选翻译，完整内容请查看原文。

Timestral: Mistral’s time-series foundation models

Timestral, Mistral’s first family of time-series foundation models, delivers state-of-the-art probabilistic, multivariate forecasts with just 209M parameters, outperforming larger models on key benchmarks while supporting past and future covariates. Available in four sizes (5M to 209M parameters), it balances accuracy and efficiency, running efficiently on CPUs and excelling in zero-shot forecasting for diverse domains. The model combines synthetic pre-training, covariate-aware attention, and long-context learning, with fine-tuning options for domain adaptation via Mistral Forge.

Timestral 是 Mistral 推出的首个时间序列基础模型系列，仅需 2.09 亿参数即可提供最先进的概率性多变量预测。它在关键基准测试中表现优于更大的模型，并支持过去和未来的协变量。该系列提供四种尺寸（500 万至 2.09 亿参数），在准确性和效率之间取得了平衡，可在 CPU 上高效运行，并在不同领域的零样本预测中表现出色。该模型结合了合成预训练、协变量感知注意力机制和长上下文学习，并可通过 Mistral Forge 进行微调以实现领域适配。

Today, we’re releasing Timestral, our first family of time-series foundation models. Our largest model, at just 209M parameters, ranks first on all three of the field’s main public benchmarks, GIFT-Eval, TIME, and fev-bench, outperforming the previous state of the art with only two-thirds of the model parameters.

今天，我们发布了 Timestral，这是我们首个时间序列基础模型系列。我们最大的模型仅有 2.09 亿参数，在 GIFT-Eval、TIME 和 fev-bench 这三个领域内主要的公共基准测试中均排名第一，仅用三分之二的模型参数就超越了之前的业界最佳水平。

Timestral delivers probabilistic, multivariate forecasts and supports past-only and known-future covariates. Our family of models sits on the Pareto front of forecasting accuracy versus model size, delivering leading performance at each size. Results have been submitted to the GIFT-Eval and TIME leaderboards.

Timestral 提供概率性多变量预测，并支持仅包含过去数据和已知未来数据的协变量。我们的模型系列处于预测准确性与模型规模的帕累托前沿，在每种尺寸下都能提供领先的性能。相关结果已提交至 GIFT-Eval 和 TIME 排行榜。

Time series is one of the most common data types in enterprise, from sensor telemetry to observability metrics. Forecasts built on it drive decisions in capacity planning, energy, supply chains, risk and maintenance.

时间序列是企业中最常见的数据类型之一，涵盖从传感器遥测到可观测性指标的各种数据。基于此构建的预测驱动着容量规划、能源、供应链、风险管理和维护方面的决策。

Industrial plants forecast process sensors against the production schedule, catch equipment drift before it fails, and plan capacity and spare parts across thousands of assets. Finance teams forecast flows, volumes, and volatility to inform positions, hedging, and treasury across thousands of accounts.

工业工厂根据生产计划预测过程传感器数据，在设备故障前捕捉其性能漂移，并规划数千项资产的容量和备件。金融团队则通过预测资金流、交易量和波动率，为数千个账户的头寸管理、对冲和资金运作提供参考。

Classical models such as ARIMA and linear regression remain well-established tools for forecasting time series. These models are usually fitted to a single series and learn only from that series’ history.

ARIMA 和线性回归等经典模型仍然是预测时间序列的成熟工具。这些模型通常针对单一序列进行拟合，且仅从该序列的历史数据中学习。

Deep learning models such as PatchTST add the capacity to learn nonlinear patterns across many related series, but they are still typically trained from scratch on each dataset, so what they learn doesn’t carry over to new domains.

PatchTST 等深度学习模型增加了跨多个相关序列学习非线性模式的能力，但它们通常仍需在每个数据集上从头开始训练，因此其学到的知识无法迁移到新的领域。

Time series foundation models bring the pre-training approach to time series. Trained on a large and diverse corpus, our model learns patterns that transfer across series, forecasting zero-shot on data it has never seen, even in domains where data is scarce.

时间序列基础模型将预训练方法引入了时间序列领域。通过在庞大且多样化的语料库上进行训练，我们的模型学习到了可跨序列迁移的模式，即使在数据稀缺的领域，也能对从未见过的数据进行零样本预测。

We evaluated Timestral on the three main public forecasting benchmarks: GIFT-Eval, TIME and fev-bench. Among pre-trained, zero-shot models, it ranks first across all five metrics. These results come from a single checkpoint, without ensembling.

我们在三个主要的公共预测基准测试（GIFT-Eval、TIME 和 fev-bench）上评估了 Timestral。在所有预训练的零样本模型中，它在全部五个指标上均排名第一。这些结果来自单一检查点，且未使用集成学习。

Timestral has been optimised as a family of four models: 5M, 28M, 123M and 209M parameters. The larger models improve forecast performance, while the smaller ones are faster and cheaper. All four share the same architecture and training recipe. Every size sits on the Pareto front of accuracy versus model size: for each model size, no public model forecasts better.

Timestral 已被优化为一个包含四种模型的系列：500 万、2800 万、1.23 亿和 2.09 亿参数。较大的模型提升了预测性能，而较小的模型则更快、成本更低。这四种模型共享相同的架构和训练方案。每种尺寸都处于准确性与模型规模的帕累托前沿：在同等模型规模下，没有其他公开模型能提供更好的预测效果。

Timestral does not need a GPU at inference. On a standard server CPU with 8 threads, forecasting one day ahead from three weeks of 15-minute history takes 10 ms with the 5M model and 140 ms with the 209M model. Smaller models trade some accuracy for speed: the 28M model is five times faster than the 209M model, and the 123M model comes within 0.2% of the 209M score at three quarters of its latency.

Timestral 在推理时不需要 GPU。在配备 8 线程的标准服务器 CPU 上，使用 500 万参数模型根据三周的 15 分钟历史数据预测未来一天仅需 10 毫秒，而 2.09 亿参数模型则需 140 毫秒。较小的模型以牺牲部分准确性换取速度：2800 万参数模型比 2.09 亿参数模型快五倍，而 12300 万参数模型的得分与 2.09 亿参数模型相差不到 0.2%，且延迟仅为其四分之三。

Consider an industrial plant with thousands of sensors across its equipment. Timestral can improve the forecasts on that data in several ways.

以一家在设备上安装了数千个传感器的工业工厂为例。Timestral 可以通过多种方式改进对这些数据的预测。

Multivariate forecasting. A pump’s vibration, temperature and pressure don’t move independently. A rise in one often explains or causes a change in another. Timestral models related series jointly, so it can use those dependencies instead of forecasting each sensor in isolation.

多变量预测。泵的振动、温度和压力并非独立变化。其中一个指标的上升往往解释或导致了另一个指标的变化。Timestral 对相关序列进行联合建模，因此它可以利用这些依赖关系，而不是孤立地预测每个传感器。

Probabilistic forecasting. A single predicted value may not show the full picture. Timestral predicts a full set of quantiles, so teams can see how wide the uncertainty is, set alert thresholds that reflect it, and plan maintenance around risk rather than a single number.

概率性预测。单一的预测值可能无法展现全貌。Timestral 预测出一整套分位数，因此团队可以观察不确定性的范围，设置反映该范围的警报阈值，并围绕风险而非单一数值来规划维护工作。

Categorical covariates. Some covariates take values from a set of categories rather than continuous numbers, such as the day of the week, or a machine’s state (running, idle, under maintenance). Encoding these as numbers would impose an order or scale they don’t have. Timestral takes them as categorical inputs directly to improve forecasts.

分类协变量。某些协变量取值于一组类别而非连续数字，例如星期几或机器状态（运行、空闲、维护中）。将这些编码为数字会强加它们本不具备的顺序或尺度。Timestral 直接将它们作为分类输入，以改进预测效果。

Known-future covariates. Some drivers of sensor behaviour are known in advance, such as production schedules, planned shutdowns, weather forecasts. Timestral can condition on them to improve forecast quality, so a scheduled load increase is expected rather than flagged as a surprise.

已知未来协变量。某些传感器行为的驱动因素是预先已知的，例如生产计划、计划停机和天气预报。Timestral 可以基于这些条件来提高预测质量，从而使计划内的负载增加被视为预期之内，而非被标记为突发异常。

Timestral is a numerical model that can run inside existing monitoring pipelines, or be paired with our language models in multi-modal agentic workflows.

Timestral 是一个数值模型，既可以在现有的监控流水线中运行，也可以与我们的语言模型配对，应用于多模态智能体工作流中。

In the plant, that could look like this:

在工厂中，其应用场景可能如下：

Timestral running continuously on the plant’s sensor streams, detecting anomalies as they occur.

Timestral 在工厂的传感器数据流上持续运行，实时检测异常情况。

When it flags an anomaly on a pump, an agent embeds the signals and compares against embeddings of past sensor data to find similar historical failures.

当它标记出泵的异常时，智能体会对信号进行嵌入，并将其与过去传感器数据的嵌入进行比较，以寻找相似的历史故障。

The agent then starts sub-agents to run a root cause analysis, drawing on context from that plant, such as maintenance records, shift logs and operator notes, equipment manuals, past work orders, and readings from rel

随后，智能体会启动子智能体进行根本原因分析，利用该工厂的背景信息，如维护记录、轮班日志、操作员笔记、设备手册、过往工单以及相关读数。