| DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression | ✓ Link | 88.1 | DeepSeek-V4.1-Flash Max (DeepSeek Harness (Minimal)) | 2026-09-17 |
| Every FLOP Counts: Scaling a 300B Mixture-of-Experts LING LLM without Premium GPUs | ✓ Link | 87.9 | Ling-3.1-flash | 2025-03-07 |
| Atria Dawn: The Dawn of Agentic Superintelligence | ✓ Link | 86.5 | Atria Dawn Preview (Claude Code v2.1.119) | 2026-09-14 |
| GLM-5.3: Frontier Coding with Emergent Cyber Capabilities | ✓ Link | 84.5 | GLM-5.3 (Claude Code) | 2026-08-14 |
| DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence | ✓ Link | 83.3 | DeepSeek-V4-Pro-0813 (Max) | 2026-04-26 |
| Hy4 preview | ✓ Link | 78.4 | Hy4 preview | 2026-08-28 |
| DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence | ✓ Link | 76.7 | DeepSeek-V4-Flash-0731 (Max) | 2026-04-26 |
| On the Limitations of Non-GPU AI Accelerators for Large-Model Inference: A Field Study of MoE and Multimodal Serving on Huawei Ascend | | 75.3 | DeepSeek-V4-Flash-Vision-Exp (DeepSeek Harness) | 2026-07-09 |
| Feyospace-v1: How the Cyber Mercury Seven Trained Frontier Cyber Models | | 63.24 | Feyospace-s1 (Claude Code 2.1.210) | 2026-09-08 |
| CyberFactory: Scaling Cyber Security Capabilities with Instances from the Wild | ✓ Link | 58.1 | OpenAegis | 2026-08-24 |
| Muse Spark Safety & Preparedness Report | | 43.5 | Muse Spark Open (Minimal ReAct agent in a CentOS CyberGym environment; one submission attempt per challenge) | 2026-05-14 |