Post by emaad khwaja

🚀 We’re thrilled to announce the open release of Toto and BOOM—two major steps forward in time series modeling for observability. 🔍 Toto is a 151M-parameter, open-weights foundation model trained entirely on real-world observability data from Datadog’s telemetry systems. It achieves state-of-the-art performance (by a wide margin!) not just on our newly released BOOM benchmark, but also on widely used datasets like GIFT-Eval and LSF. 📊 BOOM (Benchmark of Observability Metrics) is a public dataset with 350M observations across 2,800+ real production time series, designed specifically to evaluate the unique challenges of observability metrics: high cardinality, noise, sparsity, and the cold-start problem. It’s built to push the TSFM community forward. Highlights: 🧠 Zero-shot forecasting with Toto—no fine-tuning needed 🏆 Best-in-class performance on multiple benchmarks 📈 Evaluation code + leaderboard to track progress in real time Both are open source (Apache 2.0) and available now for the community to build on (Links in comments). I feel so fortunate to have worked with such a brilliant and dedicated team to bring this to life: Ben Cohen Youssef DOUBLI salahidine lemaachi Chris Lettieri Charles Masson Hugo MICCINILLI Elise Ramé Qiqi Ren Afshin Rostamizadeh Jean du Terrail Anna Monica Toon Kan Wang Stephan Xie Zongzhe Xu Viktoriya Zhukova David Asker Ameet Talwalkar Othmane Abou-Amal If this kind of work excites you, we’re hiring in NYC and Paris—join us! #opensource #ai #machinelearning #timeseries #observability #mlops #foundationmodels #datadog #toto #boom