The LLM’s deployment options include an API and local deployment on-premises. Mistral said the local deployment option could help more regulated industries, such as finance, banking, and healthcare, adopt the model.
In benchmark tests, such as Arabic MMLU, Arabic TyDiQAGoldP, Arabic Alghafa, and Arabic Hellaswag, Saba outperforms Mistral Small 3, Qwen 2.5 32B, Llama 3.1 70B, and G42’s Jais 70B.
Saba also outperforms LLama 3.3 70B Instruct, Cohere Command-r-08-2024 32B, Jais 70B Chat, and GPT-4o-mini in benchmarking tests, such as Arabic MMLU Instruct, Arabic MT-Bench Dev, and Arabic-Centric FLORES-101.
