OpenAI fixes 'lazy' GPT-4 AI model, makes GPT-3.5 Turbo cheaper
OpenAI has made exciting updates, slashing API access prices and introducing new models. For developers using ChatGPT, the GPT-3.5 Turbo—the most interacted model—is now more wallet-friendly, with input costs reduced by 50% and output by 25%. This change comes as other open source or self-managed models start to rival OpenAI's performance, pushing the company to offer competitive pricing to retain its customers.
GPT-3.5 Turbo gets 'various improvements'
GPT-3.5 Turbo has received a fresh model version (0125), which brings "various improvements." Though OpenAI hasn't shared specifics, this model will feature improved precision in generating responses in specified formats and will address bugs. GPT-3.5 Turbo has established itself as the industry standard, setting a high bar for other AI models aiming to rival OpenAI in terms of accuracy and brand reputation.
GPT-4 Turbo receives a fresh fix for 'laziness'
OpenAI is also introducing an updated version (0125) of the GPT-4 Turbo preview model, tackling an issue where the model was "lazy" in tasks like code generation. "This model completes tasks like code generation more thoroughly than the previous preview model and is intended to reduce cases of 'laziness' where the model doesn't complete a task," said the company's in its official blog post. Initially in preview mode, OpenAI plans to launch the general availability of GPT-4 Turbo soon.
OpenAI also introduced new text embedding models
There are a series of new and improved text embedding models as well. OpenAI has introduced a new version of its free moderation API, version 007, which helps identify potentially harmful text content for those looking to boost their moderation capabilities.
