TSQueryBench - Anomaly Detection F1: leaderboard

Locating anomalies in synthetic time series on TSQueryBench (2026): 100 series of length 100 to 500 hold between one and ten injected anomalies, and predicted indices count as hits within two time steps of a true anomaly. F1 in percent; higher is better.

Metric: Multi-anomaly detection F1 (%, times 100): predicted anomaly indices matched to the true ones within 2 time steps over 100 synthetic series with 1 to 10 anomalies each, zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemma 4 31B (IT)88#354
2DeepSeek V3.287#425
3Claude 3 Haiku62#2683
4Qwen 3 8B53#1890
5Llama 3.1 8B Instruct34#4655
6Gemma 3 12B (IT)28#1907

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=tsquerybench-anomaly-detection-f1 · How It Works · Data refreshed daily, snapshot 2026-11-22.