HybridAgent Evaluation - HLE Computer Engineering: leaderboard
HybridAgent Evaluation - HLE Computer Engineering. Paper agentic evaluation task subsets;five independent runs;source Table1. HybridAgent checkpoints trained separately and never aliased to base Qwen. Metric: Accuracy (%). Primary published table; source configurations remain distinct.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around January 2028. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K3 | 32 | #30 |
| 2 | GPT-5.6 Luna | 25.3 | #69 |
| 3 | DeepSeek V4.1 Flash | 25.3 | #36 |
| 4 | GLM-5.3 Flash | 14.7 | #66 |
| 5 | QwQ-32B | 7.3 | #415 |
| 6 | Qwen 3 14B (Reasoning) | 6.7 | #530 (Qwen 3 14B) |
| 7 | Qwen 3 8B (Thinking) | 2.7 | #669 (Qwen 3 8B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=hybridagent-evaluation-hle-computer-engineering · How It Works · Data refreshed daily, snapshot 2026-10-16.