Gimlet Labs joins MLCommons to develop agentic inference benchmarks
Gimlet Labs →Gimlet Labs announced its membership in MLCommons, the nonprofit consortium responsible for MLPerf benchmarks. The company said its team had already worked within the Reasoning LLM task force alongside NVIDIA, AMD and other participants on GPT-OSS-120B benchmarks and updates for interactive DeepSeek-R1 scenarios. Its next focus is developing common benchmarks for agentic inference. Gimlet described a need to compare workloads across software, networking and heterogeneous hardware using open, reproducible methods. The initiative supports its work distributing inference across different processors and architectures. This is a standards and research collaboration, with future benchmark development distinguished from work the company said it had already completed.