AARRI-Bench (Claude Code) - Context: leaderboard

Metric: Task success rate (%) under the 0/1 reward on the context tasks: sensitivity to the state of a field (core contributions, data validity, genuine advances); Harbor containerized runs, Claude Code agent harness; a configuration's tasks without a valid run are left out of its denominator; higher is better. Source: arxiv.org. Saturation forecast: Around December 2027. 6 models tracked.

Top models

#ModelScore
1Claude Opus 4.755.9
2Claude Sonnet 4.650
3Qwen 3.6 Plus50
4MiniMax-M2.747.1
5GPT-5.3 Codex47.1
6Kimi K2.645.5

Interactive version: theaggregate.ai/benchmark?slug=aarri-bench-claude-code-context · How It Works · Data refreshed daily, snapshot 2026-09-29.