Major Release OpenAI LLM Proprietary

OpenAI GPT-o1-preview

OpenAI's first public reasoning model — 'thinking before answering'

Released 2024-09-12 · ~175B · 128K tokens context · knowledge cutoff 2024-04

Overview

OpenAI's most advanced reasoning model designed for complex problem-solving

Specifications

DeveloperOpenAI
Release date2024-09-12
Model typeLLM
Parameters~175B
ArchitectureTransformer with chain-of-thought reasoning
Context window128K tokens
Max output32768 tokens
Knowledge cutoff2024-04
LicenseProprietary
Input modalitiestext
Output modalitiestext

Benchmarks

Benchmark GPT-o1-preview GPT-4Δ
AIME 83rd percentile
GPQA 78%
MATH 83% 52.9% +30.1

Availability

API Only OpenAI API Azure OpenAI

What's new vs GPT-4

Direct comparison available in the model comparison tool →

Official references