One of the great strengths of ChatGPT is its ability to aid in creative writing. ChatGPT’s latest large language model, GPT-4o, has received a bit of a performance boost, OpenAI announced Wednesday. Users can reportedly expect “more natural, engaging, and tailored writing to improve relevance & readability” moving forward.
GPT-4o got an update 🎉
The model’s creative writing ability has leveled up–more natural, engaging, and tailored writing to improve relevance & readability.
It’s also better at working with uploaded files, providing deeper insights & more thorough responses.
— OpenAI (@OpenAI) November 20, 2024
GPT-4o, not to be confused with 01 (formerly Project Strawberry), is OpenAI’s latest publicly available model, surpassing GPT-4 and GPT-3.5. GPT-4o was first released in May 2024 and offers users double the performance at half the resource cost of its direct predecessor, GPT-4-Turbo, as well as state-of-the-art benchmark results in voice, multilingual, and vision tasks. Not only is it more efficient than older versions but it offers a host of additional capabilities. The model’s rapid response pace makes it especially useful for real-time translation and conversation applications.
For example, ChatGPT’s Advanced Voice Mode would not be possible without the near-instantaneous text, voice, and audio inference that GPT-4o offers (and that previous models did not). GPT-4o boasts improved reasoning skills over its predecessors as well, able to interpret a user’s spoken tone and intention, understand unique aspects of their tone, pace, and mood, and reply in kind.
While GPT-4o is technically available to every OpenAI subscriber level, including the free tier, it is not available for unlimited use. Free-tier users will find that they can only access the new model a handful of times on ChatGPT before being shunted down to interact with the smaller GPT-4o-mini model. Plus, Teams, and Enterprise subscribers have a rate limit roughly five times higher.
GPT-4o-mini is based on the same training data as its larger sibling, but leverages fewer variables (not to mention fewer compute resources) in its inference operations. Being lighter-weight and more responsive, GPT-40-mini has proven useful in a variety of small-scale applications including computer code generation.
OpenAI claims GPT-4o mini is “the most capable and cost-efficient small model available today,” per CNBC, out-performing competitors like Google’s Gemini 1.5 Flash, Meta’s Llama 3 8b, or Anthropic’s Claude 3 Haiku in a variety of benchmarks. According to data from Artificial Analysis, 40-mini scored 82% on the MMLU reasoning benchmark, topping Gemini by 3% and Claude 3 Haiku by 7%.