GPT-5.1 Thinking

GPT-5.1 Thinking is the next iteration of GPT-5, designed with enhanced adaptive reasoning capabilities. Unlike GPT-5.1 Instant, GPT-5.1 Thinking adapts thinking time more precisely to each question, providing more thorough reasoning for complex queries. It offers improved conversational abilities with clearer and more engaging interactions, reducing jargon and enhancing user comprehension. GPT-5.1 Thinking demonstrates significant improvements in coding tasks and complex reasoning scenarios, making it valuable for developers and users who need deeper analysis.

Benchmark results

Benchmark Score Tags Source
AIME 2025 94.0% self-reported llm-stats link →
BrowseComp Long Context 128k 90.0% self-reported llm-stats link →
FrontierMath 26.7% self-reported llm-stats link →
GPQA 88.1% self-reported llm-stats link →
MMMU 85.4% self-reported llm-stats link →
SWE-Bench Verified 76.3% self-reported llm-stats link →
Tau2 Airline 67.0% self-reported llm-stats link →
Tau2 Retail 77.9% self-reported llm-stats link →
Tau2 Telecom 95.6% self-reported llm-stats link →