Cisco Released Cisco Time Series Model: Their First Open-Weights Foundation Model based on Decoder-only Transformer Architecture

Cisco Time Series Model

Cisco and Splunk introduced the Cisco Time Series Model, an open-weight foundation model that reduces mean absolute error by 25% on observability benchmarks. It extends TimesFM 2.0 with a multiresolution architecture, handling coarse and fine-grained data in one context window.

Why This Matters

Traditional time series models operate on single-resolution data, but real-world observability metrics require both long-term trends (coarse) and short-term spikes (fine). Poor resolution handling can lead to missed anomalies or inaccurate forecasts, costing enterprises in downtime or security breaches. Cisco’s model addresses this by explicitly fusing coarse and fine data, improving accuracy on 1-minute observability workloads.

Key Insights

• “300B+ data points in training, 40% from observability (Splunk)”
• “Multiresolution context: coarse (512 hours) + fine (512 mins) for 1-minute forecasts”
• “Cisco’s model outperforms Chronos 2 and AutoARIMA on observability benchmarks”

Practical Applications

• Use Case: Observability platforms using Cisco’s model for real-time anomaly detection in 1-minute metrics
• Pitfall: Over-reliance on coarse data may miss short-term spikes if fine-grained data is incomplete

References:

• https://www.marktechpost.com/2025/12/07/cisco-released-cisco-time-series-model-their-first-open-weights-foundation-model-based-on-decoder-only-transformer-architecture/