Measuring how well CLI agents like Claude Code or Codex CLI can post-train base LLMs on a single H100 GPU in 10 hours
2 files for Codex, OpenCode and Claude Code: PostTrainBench AGENTS.md, PostTrainBench CLAUDE.md — 8,547 tokens loaded in every session.
AGENTS.md D 8,543 tok CLAUDE.md A 4 tok These files are aisa-group/PostTrainBench's own configuration — they tell Codex, OpenCode and 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.