meshllm/performance-history

MeshLLM Performance History Append-only, machine-readable results from MeshLLM's trusted nightly competitive benchmark. Each immutable run shard contains one normalized row per backend, model, concurrency, hardware/config cohort, and source revision. The dataset contains performance metrics and content-addressed provenance only. It does not contain prompts, completions, model weights, credentials, local filesystem paths, or raw benchmark logs. GitHub Actions retains the… See the full description on the dataset page: https://huggingface.co/datasets/meshllm/performance-history.