TALES: leaderboard

Metric: Score (% of max). Source: microsoft.github.io. 39 models tracked.

Top models

#ModelScore
1O3 (Medium)58.7
2O3 (High)58
3O3 (Low)54.8
4Claude 3.7 Sonnet (Thinking)52.5
5Claude 3.7 Sonnet52.1
6GPT-4.149.9
7O144.2
8GPT-4o40.6
9Claude 3.5 Haiku39.6
10Llama 3.1 405B Instruct36.4
11Gemini 2.0 Flash35
12Qwen 3 32B34.3
13Llama 3.3 70B Instruct32.8
14Llama 3.1 70B Instruct32
15Qwen 2.5 72B Instruct30.7

Interactive version: theaggregate.ai/benchmark?slug=tales · How It Works · Data refreshed daily, snapshot 2026-09-19.