Chronos-2
A publicly available foundation model for time-series forecasting that can be adapted to specific dataset configurations via LoRA or full fine-tuning.
Timeline (last 90 days)
Timestral-209M scores 0.658 MASE on GIFT-Eval, versus 0.661 for t0-gamma, 0.667 for TimesFM-3, and 0.698 for Chronos-2.
Timestral-209M records 50.39 on the fev-bench SQL skill score, versus 49.42 for t0-gamma and 47.28 for Chronos-2.
Saudi Arabia's Fodder Watch submitted two models built on fourteen objects: a detector RF-DETR (Apache 2.0 license, fine-tuned by region) and a predictor Chronos-2 (Apache 2.0 license, zero-shot, approximately 120 million parameters).
Chronos-2 is a time-series forecasting foundation model developed by Amazon, released under Apache 2.0, with approximately 120 million parameters, and supports zero-shot time-series forecasting.
TW3Cast applies LoRA or full fine-tuning to three publicly available time series foundation models: Chronos-2, TiRex, and Toto.
All Facts (5)
Timestral-209M scores 0.658 MASE on GIFT-Eval, versus 0.661 for t0-gamma, 0.667 for TimesFM-3, and 0.698 for Chronos-2.
50%UnverifiedTimestral-209M records 50.39 on the fev-bench SQL skill score, versus 49.42 for t0-gamma and 47.28 for Chronos-2.
50%UnverifiedSaudi Arabia's Fodder Watch submitted two models built on fourteen objects: a detector RF-DETR (Apache 2.0 license, fine-tuned by region) and a predictor Chronos-2 (Apache 2.0 license, zero-shot, approximately 120 million parameters).
50%UnverifiedChronos-2 is a time-series forecasting foundation model developed by Amazon, released under Apache 2.0, with approximately 120 million parameters, and supports zero-shot time-series forecasting.
50%UnverifiedTW3Cast applies LoRA or full fine-tuning to three publicly available time series foundation models: Chronos-2, TiRex, and Toto.
50%