OpenAI launches new generation of integration models and other API updates

OpenAI, THE artificial intelligence research business, announcement on THURSDAY A new generation of integration models, which can convert text In A digital form that can be used For miscellaneous machine learning Tasks. THE business Also introduced new variants of It is GPT-4 Turbo And moderation models, new API use management tools, And lower price on It is GPT-3.5 Turbo model.

Integrations are sequences of Numbers that represent THE Notions In content such as natural language Or coded. Integrations TO DO he easy For machine learning models And other algorithms has to understand THE relationships between content And has perform Tasks as grouping Or recovery. They power apps as awareness recovery In both ChatGPT And THE Assistants API, And a lot recovery increase generation (CLOTH) developer tools.

OpenAI said that It is new integration models, text-3-small embedding And text-3-large embedding, offer stronger performance And reduced price compared with has It is previous generation model, text-integration-ada-002. THE new models can create integrations with up has 3072 dimensions, which can capture more semantics information And improve THE precision of downstream tasks.

According to has THE business, THE new models to have increase THE average score on A commonly used reference For multilingual recovery (MIRACLE) Since 31.4% has 54.9%, while THE average score on A commonly used reference For English Tasks (MTEB) has increase Since 61.0% has 64.6%. THE price For text-3-small embedding has Also has been reduced by 5X compared with has text-integration-ada-002, manufacturing he more affordable For developers has use.

THE business Also update It is GPT-4 Turbo And GPT-3.5 Turbo models, which are big multimodal models that can to understand And generate natural language Or coded. THE new variants of THE models come with improved instruction following, JSON fashion, more reproducible outputs, And parallel function call. THE business Also spear A new 16k context version of GPT-3.5 Turbo, which can process longer contributions And outputs that THE standard 4k version.

In addition, THE business update It is text moderation model, which can detect if text can be sensitive Or unsafe. THE new version of THE model can handle more LANGUAGES And areas, And can Also provide explanations For It is predictions.

THE business Also introduced new manners For developers has manage API keys And to understand API use. Developers can NOW create several API keys with different permissions And the scopes, And monitor their use And billing details on THE OpenAI Dashboard. THE business Also said that he will Soon lower THE price on It is GPT-3.5 Turbo model by 25%, manufacturing he more accessible For developers has build apps with that.

OpenAI said that these updates are part of It is continuous efforts has improve It is models And services, And has TO DO them more useful And affordable For developers And clients. THE business Also guest developers has contribute evaluations has help he improve THE model For different to use case. THE business said that he will continue has release new models, features, And tools In THE future.

VentureBeat's assignment East has be A digital city square For technical decision-makers has earn awareness about transformer business technology And to treat.