ARFBench - End Time: leaderboard

Time series question answering on software incident data in ARFBench (2026, Datadog): multiple-choice questions over multivariate production telemetry from real incidents, with the series shown as chart images. Accuracy on the End Time questions (when the last anomaly ends), in percent; higher is better.

Metric: Accuracy (%) on the 32 End Time questions (when the last anomaly ends, or no anomaly) of ARFBench, for the time series rendered as chart images (Matplotlib plots, three images for paired questions), with the series description and channel names as text; few-shot multiple-choice prompting with shuffled answer choices, temperature 0.05 (default for models without the setting), reasoning models at medium effort with 2,000 output tokens, blank or invalid answers counted wrong; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-568.8#185
2GPT-5.4Medium65.6#138
3Gemini 3 Pro62.5#146
4Toto-1.0-QA-Experimental 32B (TSFM-VLM)59.4
5Qwen 3 VL 32B56.3#624
6Claude Opus 4.653.1#114
7GPT-4o43.8#754
8Qwen 3 VL 32B (post-trained)43.8
9GPT-4.140.6#506
10Claude Sonnet 4.537.5#267
11Qwen 3 VL 8B31.3#927

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

Interactive version: theaggregate.ai/benchmark?slug=arfbench-end-time · How It Works · Data refreshed daily, snapshot 2026-11-22.