Tech Desk 
Jul 22, 2024
Apple is rapidly expanding its ecosystem in the field of artificial intelligence (AI) and offering the latest 7B parameter model which would be accessible for everyone to adopt.
Recently, the Cupertino-based tech giant unveiled the DataComp for Language Models (DCLM), surpassing the Mistral-7B which is ultimately approaching same-sized models from Meta and Google.
Vaishaal Shanker — an Apple employee— wrote on X (formerly Twitter) that Apple was the"best performing truly open-source model" now.
These remarks demonstrate that the truly open-source AI model comes with all the weights, datasets, training code, and more along with the model.
Read more: Rumour points towards AI deal between Apple, Meta for iOS 18
It came under the light of development in the same week when Meta is likely to launch GPT-4 competitor Llama 3 400B. However, the Cupertino-based tech giant’s further plans encompassing the larger DCLM model shortly are still under wraps.
What you must know about Apple’s AI model?
Apple has been aiming bigger and constantly working to accomplish its goals to achieve high-quality design of datasets for the training models.
Moreover, the entire team diligently ran multiple experiments across similar model architecture, evaluations, training codes, and more to figure out the best data strategy to implement in their plan and drive insightful results that perform efficiently.
Apple received fruitful results for their constant efforts which resulted in DCML-Baseline, which was supposed to train the latest models in seven billion and 1.4 billion parameter versions.
Additionally, this 7B fully open-source AI model truly depicts efficiency in the training data and processes, making it one of the most crucial AI releases of the year.
Sam Altman, OpenAI chief executive officer (CEO) said that the release of the smaller GPT-4o mini aims to craft intelligence too cheap to metre, the Cupertino-based tech giant’s model is one of the best examples of it.
