High-performance LLM evaluation framework with parallel API calls — up to 17× faster than sequential tools. Supports box, math, and logit-based evaluation.
Latest release v2.10.0 — v2.10.0 — VisTW 繁體中文視覺評測 · 15 Sept 2026
3 files for Claude Code: Eval CLAUDE.md, add-benchmark, run-eval — 12,871 tokens loaded in every session.
CLAUDE.md A 12,871 tok .claude/skills/add-benchmark/SKILL.md A 151 tok .claude/skills/run-eval/SKILL.md A 129 tok These files are ai-twinkle/Eval's own configuration — they tell Claude Code how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.