OpenAI, the research organization behind some of the most advanced artificial intelligence models, has announced a major update to its flagship text-generating model, GPT-4. The new version, called GPT-4 Turbo, promises to deliver better performance, lower cost, and more up-to-date knowledge than its predecessor.
GPT-4 Turbo was unveiled at the OpenAI DevDay event, where company CEO Sam Altman shared some of the new features and improvements of the model. According to Altman, GPT-4 Turbo can process a much larger amount of text than GPT-4, thanks to its increased context window of 128,000 tokens, which is equivalent to about 96,000 words. This means that GPT-4 Turbo can handle longer and more complex tasks, such as generating novels, essays, or code, without losing track of the topic or the details¹.
Another advantage of GPT-4 Turbo is that it has a more recent knowledge base, with a cutoff date of April 2023. This means that the model can access and generate information that is more relevant and accurate for the current world, such as the latest news, events, or trends¹. For example, GPT-4 Turbo can answer questions about the recent nuclear fusion experiment from South Korea, or the upcoming Cricket World Cup, which GPT-4 may not be able to do.
Additionally, GPT-4 Turbo offers a more cost-effective pricing structure, with a lower price per token and a higher token rate limit. This means that developers and customers can use the model more frequently and affordably, without compromising on the quality or the speed of the output². Altman also announced that OpenAI is launching a fine-tuning program for GPT-4, which will allow users to customize the model for their specific use cases and domains, such as education, health, or entertainment³.
GPT-4 Turbo is available in two variants: one that can understand and generate text or code, and another that can also comprehend images. The latter variant is powered by DALL-E, a model that can generate and edit images given a natural language prompt². Both variants of GPT-4 Turbo are accessible through the OpenAI API, which also provides access to other models, such as TTS, Whisper, Embeddings, and Moderation².
OpenAI is a research organization that aims to create and promote artificial intelligence that can benefit humanity, without causing harm or discrimination. The company is known for its groundbreaking work on large language models, such as GPT-3 and GPT-4, which can generate realistic and diverse texts on almost any topic. OpenAI also develops multimodal models, such as DALL-E and CLIP, which can combine vision and language to create and understand images².
(1) OpenAI introduces GPT-4 Turbo: Larger memory, lower cost, new knowledge. https://arstechnica.com/information-technology/2023/11/openai-introduces-gpt-4-turbo-larger-memory-lower-cost-new-knowledge/.
(2) OpenAI Platform. https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo.
(3) OpenAI Unveils GPT-4 Turbo And Fine-Tuning Program For GPT-4. https://robots.net/news/openai-unveils-gpt-4-turbo-and-fine-tuning-program-for-gpt-4/.