Harbor-Index 1.0 (Claude Code): leaderboard
Harbor-Index 1.0's 82 hard agentic tasks, curated from 29 Harbor-ported benchmarks in software engineering, science, tool use, knowledge, mathematics, data analytics and security, run in the Claude Code agent: Claude Opus 4.8 through Anthropic's API and the other models through OpenRouter. The score is the pass rate in percent over the 82 tasks, one run per task with default model settings, higher is better; continuous speedup tasks pass only at a locked threshold set by the best observed result, and tasks were kept only where Claude Opus 4.6, GPT-5.4 and Gemini 3.1 Pro together solved at most a third of 18 trials (Harbor framework team, 2026).
Metric: Pass rate (%; 82 tasks, one run each). Source: arxiv.org. Saturation forecast: Around November 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.8 | 20.7 | #56 |
| 2 | GLM-5.2 | 8.5 | #137 |
| 3 | Kimi K2.6 | 6.1 | #173 |
| 4 | DeepSeek V4 Pro | 4.9 | #181 |
| 5 | Qwen 3.7 Max | 4.9 | #142 |
| 6 | MiniMax-M3 | 3.7 | #265 |
| 7 | MiMo-V2.5-Pro | 2.4 | #274 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=harbor-index-1-0-claude-code · How It Works · Data refreshed daily, snapshot 2026-11-12.