Guide

GPT-4o mini

GPT-4o mini is the light version of GPT-4o: most of the everyday competence at a fraction of the cost. It is the budget pick of the OpenAI lineup on Magnus.

← All guides

At a glance

Launched July 18, 2024 by OpenAI, replacing GPT-3.5 Turbo as the default small model. It kept multimodal input (text and images) while cutting price to a level that made high-volume use routine.

On the All-inclusive plan a GPT-4o mini reply costs 1 credit.

Benchmarks

  • MMLU: 82.0% - at release, the strongest score ever posted by a model in its price class.
  • Outscored GPT-4 on chat preference in the LMSYS arena leaderboard at launch.

Best at on Magnus

Quick questions, summaries, casual chat, and secondary agents that talk a lot but do not need deep reasoning. A good pick for a high-energy persona that fires off short replies all day.

  • For the same 1 credit, GPT-5.6 Luna is newer and generally stronger - mini remains the pick if you specifically want the 4o family's style.