Skip to main content
ModelScale

Agentic benchmark

InferenceBench leaderboard

Every model the catalog carries a published InferenceBench value for, ranked by that value.

CategoryAgentic
MeasureTwo-hour autonomous CLI agent run
Tasks4 inference-serving optimization scenarios
DifficultyOpen-ended ML systems engineering
Published byInferenceBench

A benchmark for open-ended LLM inference optimization by AI agents. Agents receive a base model, one H100, and a fixed time budget to build a valid OpenAI-compatible inference server that improves serving speed.

No model in the catalog has a published InferenceBench score.

The benchmark is defined by InferenceBench, but the catalog carries no value for it yet. An absent value is shown as absent here rather than as a zero.

All leaderboards · Agentic capability leaderboard