TimeSage-MT - Numerical Accuracy: leaderboard

Metric: Numerical accuracy (%): typed rule-based scorer comparing stated values with code-derived gold targets (MASE threshold, F1, 1-NRMSE, ARI and similar, normalized to 0-100), averaged over the applicable tasks; Code-Enabled Reasoning setting (the model writes Python, a sandbox runs it, the model answers from stdout; 240 multi-turn time-series analysis tasks over 8 domains), temperature 0, reasoning effort high where supported and medium for GPT-5.3-Codex; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 6 models tracked.

Top models

#ModelScore
1GLM-5.118.17
2GPT-5.3 Codex17.7
3Claude Sonnet 4.617.68
4Qwen 3.5 122B A10B15.75
5MiniMax-M2.715.54
6Gemini 3 Flash (Preview)15.25

Interactive version: theaggregate.ai/benchmark?slug=timesage-mt-numerical-accuracy · How It Works · Data refreshed daily, snapshot 2026-09-29.