Skip to content

Discrete Reasoning Over Paragraphs

A reading-comprehension benchmark requiring discrete reasoning over paragraphs, reported in DeepSeek-V4 base-model evaluations.

5 models

Higher is better.

  1. 1o1-miniOpenAI83.9
  2. 2GPT-4.5OpenAI83.4
  3. 3Claude 3.5 HaikuAnthropic83.1
  4. 4Celeris-1Celeris81.4
  5. 5Soofi S 30B-A3BSoofi Project66.5
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
F1
Tasks
Paragraph reasoning questions
Authors
DeepSeek-AI (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.