LLM2024-12-20

o3

OpenAI's o3 is a reasoning model optimized for math, science, and coding.

Overview

OpenAI's o3 is the latest evolution of their reasoning models, succeeding o1. It uses massive scale reinforcement learning to solve the world's hardest problems in math and science.

Unique Factor

Superhuman performance on specific reasoning benchmarks like the AIME.

Key Capabilities

●

Deep reasoning

●

RL-optimized

●

High coding performance

Benchmarks

MMLU Score

90.5%

HumanEval (Coding)

95%

GPQA Diamond

87%

MATH Benchmark

92.1%

Top Use Cases

Scientific Discovery

Formulating hypotheses and designing experiments.

Example: “Analyze this genetic sequence and suggest potential CRISPR targets.”

Learn to Master This Model

Take our free structured o3 course — from basics to advanced techniques.

ChatGPT Course →

Technical Specs

Context128,000 tokens

Paramsunknown

LicenseProprietary

ArchReasoning / CoT

API Pricing

$15 / 1M input tokens

Output: $60 / 1M tokens

Access API

Developer

OpenAI

Creator of ChatGPT, GPT-4o, and o3 — the world's most widely used AI platform.

Prompt Library

Browse Coding Prompts →

📋

Previous Version

O1 →