The paper introduces GRANITE-20B-FUNCTIONCALLING, an open-source function-calling model trained using multi-task learning on granular tasks such as nested function calling, function chaining, parallel functions, function name detection, parameter-value pair detection, next-best function, and response generation. The model is designed to enhance large language models (LLMs) by enabling them to interact with external tools and APIs to complete complex tasks. The evaluation of GRANITE-20B-FUNCTIONCALLING on multiple out-of-domain datasets, including the Berkeley Function Calling Leaderboard and academic benchmarks, shows that it performs well or better than other open and proprietary models, particularly in terms of generalizability and hallucination rates. The model's ability to handle granular tasks and its performance on various metrics make it a robust solution for function-calling applications.The paper introduces GRANITE-20B-FUNCTIONCALLING, an open-source function-calling model trained using multi-task learning on granular tasks such as nested function calling, function chaining, parallel functions, function name detection, parameter-value pair detection, next-best function, and response generation. The model is designed to enhance large language models (LLMs) by enabling them to interact with external tools and APIs to complete complex tasks. The evaluation of GRANITE-20B-FUNCTIONCALLING on multiple out-of-domain datasets, including the Berkeley Function Calling Leaderboard and academic benchmarks, shows that it performs well or better than other open and proprietary models, particularly in terms of generalizability and hallucination rates. The model's ability to handle granular tasks and its performance on various metrics make it a robust solution for function-calling applications.