Published local LLM observations with their recorded hardware, runtime, workload and count/timing method. Historical values have broad, unqualified applicability. A method check does not establish performance for another configuration or qualify a forecast.
Snapshot baked 2026-06-16 · 0 published results
Compiled by Ben Racicot, Founder & Lead Engineer, ModelPiper — from published ToolPiper benchmark observations.
No community observations have been published in this snapshot. The model fit checker shows labelled estimates where supported, with unavailable forecasts identified.