QuantEval - Quantitative Reasoning (No CoT): leaderboard

Metric: Accuracy (%; direct answer, no chain-of-thought). Source: arxiv.org. Saturation forecast: Around 2033. 13 models tracked.

Top models

#ModelScore
1Claude Sonnet 4.544.6
2Qwen 3 30B A3B36.5
3Qwen 3 14B33.9
4Qwen 3 4B33.9
5GPT-533.8
6DeepSeek R1 Distill Qwen 14B32.3
7Qwen 3 8B29.2
8DeepSeek R1 Distill Qwen 1.5B27.9
9DeepSeek-R1-Distill-Qwen-7B26.2
10Gemini 2.5 Pro23
11DeepSeek R1 Distill Llama 8B10.4

Interactive version: theaggregate.ai/benchmark?slug=quanteval-quantitative-reasoning-no-cot · How It Works · Data refreshed daily, snapshot 2026-09-25.