Last week, OpenAI launched GPT-4.5, its largest cutting-edge model to date. OpenAI claims that GPT-4.5 is the most knowledgeable model to date, built by further extending the pre-training process. Compared with OpenAI’s older models, in addition to having more knowledge, the GPT-4.5 model also has improved writing skills and has a more refined personality.

Today, the GPT-4.5 model debuted on ChatbotArena, ranking first in most categories. In the following categories, GPT-4.5 ranks first and is far ahead in the "Multi-Turn" category. GPT-4.5 is also far ahead in the style control rankings.

  • Multiple turns

  • Hard reminder

  • programming

  • math

  • creative writing

  • Tracking guidance

  • longer queries

    • xAI's latest Grok-3 model (grok-3-preview-02-24) also appeared on the Arena leaderboard for the first time, ranking first in "Difficulty Tips" (English) and tied for first in "Coding", "Math", "Creative Writing", "Instruction Tracking" and "Long Queries". The rapid progress demonstrated by GPT-4.5 and Grok-3 highlights the increasingly fierce competition in the field of artificial intelligence.

    OpenAI’s GPT-4.5 topped several other artificial intelligence benchmarks. It ranked first in the "Knockout" benchmark. The Knockout Tournament is a multi-party competition that tests LLM's social reasoning, strategy, and deception abilities. In IQ test score rankings, GPT-4.5 outperforms all other non-inference models in the industry. On the SimpleQA hallucination rate benchmark, GPT-4.5 scores the lowest of all OpenAI’s large language models.

    Last month, OpenAI CEO Sam Altman revealed that GPT-4.5 is OpenAI’s last non-thinking chain model. Additionally, OpenAI will no longer release o3 as a standalone model. Instead, OpenAI will unify o-series and GPT-series models, creating a system that determines appropriate think times based on user queries.

    SamAltman also confirmed that even ChatGPT free tier users can use GPT-5, but only on standard smart settings. ChatGPTPlus users will be able to run GPT-5 at a higher intelligence level, while Pro users will be able to run GPT-5 at a higher intelligence level. In addition, unified mode will support all existing ChatGPT features such as voice, canvas, search, in-depth research, etc.