Flame Machine Learning - Analyze: leaderboard

Metric: Accuracy (%; the paper's Bloom-level weighted accuracy on the Analyze 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)100
2Claude Opus 4.690
3GPT-5.486
4GPT-4.1 Mini85
5Gemini 2.5 Flash Lite76
6Qwen 3 32B76
7DeepSeek R1 Distill Qwen 32B63
8Claude Haiku 4.561
9DeepSeek R1 Distill Qwen 14B48
10Qwen 3 8B47
11Gemma 3 27B (IT)43
12Gemma 3 12B (IT)33

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