Jais 2: MBZUAI, Cerebras, and Inception Release 70B-Parameter Arabic-Centric Open LLM
Summarized by AI from reporting by ArXiv cs.CL, published under our editorial policy.
MBZUAI, Cerebras, and Inception have released Jais 2, a family of open large language models optimized for Arabic. The family includes a 70-billion-parameter model, the largest open Arabic-centric LLM trained from scratch, and an 8-billion-parameter variant. Both models use a custom Arabic-centric vocabulary for efficient training and inference.
Key takeaways
- Jais 2 is a family of Arabic-centric large language models developed by MBZUAI, Cerebras, and Inception.
- The largest model in the Jais 2 family has 70 billion parameters, making it the largest open Arabic-centric model trained from scratch.
- Jais 2 models use a custom Arabic-centric vocabulary for efficient training and inference.
- The 8-billion-parameter Jais 2 model is competitive among other open models of similar size.
Researchers from MBZUAI, Cerebras, and Inception have released Jais 2, a new family of large language models designed to excel in Arabic language tasks. The family includes models with 70 billion and 8 billion parameters, making them some of the largest and most capable Arabic-centric models available.
Jais 2 Models and Their Arabic-Centric Design
Jais 2 is a set of large language models trained specifically to understand and generate Arabic text. The models are designed to perform well on a variety of tasks, such as translation, summarization, and question-answering, all within the context of the Arabic language. The family includes models of different sizes, with the largest being a 70-billion-parameter model, which is, to the best of the researchers' knowledge, the largest open Arabic-centric model trained from scratch.
Performance and Custom Vocabulary
The Jais 2 family includes two main models: a 70-billion-parameter model and an 8-billion-parameter model. Both models have been trained using a custom Arabic-centric vocabulary, which helps them to be more efficient and accurate in their language tasks. The models have been evaluated on a range of benchmarks, and they have shown strong performance across Arabic and culturally grounded tasks. The 8-billion-parameter model is also competitive among other open models of similar size.
Why Jais 2 Matters for Arabic Language AI
Jais 2 is significant because it provides a powerful tool for Arabic language processing. For people who speak Arabic, this means better translation services, more accurate search results, and improved language understanding in AI applications. For researchers and developers, it offers a robust platform for building Arabic-centric applications, from chatbots to educational tools.
Accessing Jais 2 on Hugging Face
If you're interested in trying out Jais 2, you can access the models through the Hugging Face platform. Simply search for 'Jais 2' on Hugging Face and follow the instructions to start using the models in your own projects. This is a great opportunity to explore the capabilities of these Arabic-centric models and see how they can enhance your work or research.
Frequently asked
- Is Jais 2 available for public use?
- Yes, Jais 2 models are available for public use and can be accessed through the Hugging Face platform.
- What tasks can Jais 2 perform?
- Jais 2 models are designed to perform a variety of tasks, including translation, summarization, and question-answering in Arabic.
- How does Jais 2 compare to other Arabic-centric models?
- The 8-billion-parameter Jais 2 model is competitive among other open models of similar size, and the 70-billion-parameter model is the largest open Arabic-centric model trained from scratch.