IBM released llm-d on September 8, 2026, offering a new way to optimize AI workloads without new hardware. It is the company's first hybrid cloud update since 2024.
IBM reported llm-d achieves a 30% improvement in inference speed compared to its predecessor, measured on the MLPerf benchmark. That compares with a 15% improvement from the previous version.
llm-d is built on IBM's hybrid cloud platform and targets enterprises looking to scale AI workloads efficiently. Availability begins in Q4 2026, initially for large-scale organizations.
"llm-d makes the most of the hardware you already have," said Peter Hess, AI lead at IBM. "It enables companies to run complex AI models without the need for new infrastructure."
The announcement follows IBM's recent focus on hybrid cloud solutions. IBM said the release reflects its commitment to making AI more accessible and efficient.
IBM did not say when llm-d will be available for smaller businesses, and raised the question of how well it will perform on legacy systems. The company said it will release more details in the coming months.
Source: ibm
