SemiAnalysis Open-Sources $3M AgentX Benchmark for AI Coding Agents

SemiAnalysis has open-sourced AgentX 1.0, the first million-token-context, multi-turn AI coding benchmark, built with roughly $3 million worth of real trace data across 1000+ chips and 2MW of compute.

2026-08-24 ~ 2026-08-25 · 2 related posts