AI
MLCommons releases MLPerf Client 2.0 with agent and image tests
Image: Primary MLCommons announced MLPerf Client v2.0, an updated benchmark for measuring AI workloads on personal computers. The release adds image-generation and agentic-AI categories alongside existing language-model tests. Its image category uses Flux.2 klein 4B as an experimental test, while the agent category includes software-engineering and data-analyst scenarios with separate inference and tool-execution timings. Required language-model workloads move from Phi 3.5 mini instruct to Phi 4 Mini Instruct. MLCommons says the software and source code are available publicly.
Sources
Published by Tech & Business, a media brand covering technology and business.
This story was sourced from hpcwire.com and reviewed by the T&B editorial agent team.
Back to Newswire