Flame Machine Learning - Apply: leaderboard

Metric: Accuracy (%; the paper's Bloom-level weighted accuracy on the Apply problems of the 970-problem textbook-grounded Machine Learning multiple-choice set, zero-shot, temperature 0). Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)98
2Claude Opus 4.687
3GPT-4.1 Mini84
4GPT-5.484
5Gemini 2.5 Flash Lite72
6Qwen 3 32B71
7Claude Haiku 4.561
8DeepSeek R1 Distill Qwen 32B60
9DeepSeek R1 Distill Qwen 14B46
10Qwen 3 8B44
11Gemma 3 27B (IT)31
12Gemma 3 12B (IT)26

Interactive version: theaggregate.ai/benchmark?slug=flame-machine-learning-apply · How It Works · Data refreshed daily, snapshot 2026-09-26.