| nocot-bench (NCRI) | 159.0 | | #1 of 185 | Official leaderboard |
| Molecular Biology Capabilities Test | 63.9% | | #2 of 90 | Official leaderboard |
| KORA | 73.2% | | #1 of 42 | Official leaderboard |
| SRE-Bench | 56.9% | | #1 of 35 | Official leaderboard |
| ExploitBench | 100% | | #1 of 21 | Lab model card |
| DrugDiscoveryBench | 68.7% | | #1 of 20 | Official leaderboard |
| ExploitGym | 42.4% | | #1 of 18 | Official leaderboard |
| BrowseSafe++ | 91.5% | | #1 of 15 | Official leaderboard |
| World-Class Bio | 62.6% | | #5 of 74 | Official leaderboard |
| VIALS 2 | 49.9% | | #1 of 14 | Official leaderboard |
| Biopharma Bench | 63% | | #1 of 13 | Official leaderboard |
| VCT-v2 | 58% | | #2 of 25 | Official leaderboard |
| Human Pathogen Capabilities Test | 69% | | #7 of 86 | Official leaderboard |
| FORTRESS | 12.1% | | #7 of 73 | Official leaderboard |
| DeepsecBench | 37.6 | | #3 of 29 | Official leaderboard |
| SecIT Bench | 81% | | #3 of 26 | Official leaderboard |
| Declining to Exploit a Honeypot During Difficult ExploitGym Problems | 0% | | #1 of 7 | Lab model card |
| Instruction hierarchy attacks | 99.99% | | #1 of 7 | Lab model card |
| Prompt injection robustness | 99.8% | | #1 of 7 | Lab model card |
| Respecting Auto-Review | 0% | | #1 of 7 | Lab model card |
| BioSecBench-Function | 48.6% | | #4 of 27 | Official leaderboard |
| SpatialBench-Long | 39.3% | | #4 of 27 | Official leaderboard |
| TroubleshootingBench | 48.4% | | #3 of 20 | Lab model card |
| CWE-bench | 68% | | #2 of 12 | Official leaderboard |
| WellnessBench | 64% | | #2 of 12 | Official leaderboard |
| SHP2 Protein Function Prediction | 0.40 | | #1 of 6 | Lab model card |
| VariantBench | 46.6% | | #5 of 27 | Official leaderboard |
| TxBench-Oligo-Discovery | 51.9% | | #6 of 30 | Official leaderboard |
| EpiBench | 28.3% | | #6 of 27 | Official leaderboard |
| Multimodal Troubleshooting Virology | 55.8% | | #6 of 27 | Lab model card |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 64.6% | | #16 of 71 | Official leaderboard |
| MetagenomicsBench | 49.7% | | #6 of 26 | Official leaderboard |
| TxBench-Antibody-Discovery | 56% | | #7 of 30 | Official leaderboard |
| PCT (Polarized Contrastive Pairs) | 37.4% | | #11 of 46 | Official leaderboard |
| BioSimBench | 0.16 | | #3 of 12 | Official leaderboard |
| Suicide Memory Bench | 1.15 | | #3 of 12 | Official leaderboard |
| Sandbox Bench | 45.5% | | #1 of 4 | Lab model card |
| ProtocolQA Open-Ended | 41.4% | | #7 of 27 | Lab model card |
| scBench-Long | 27.3% | | #7 of 27 | Official leaderboard |
| SEC-bench Pro | 85.4% | | #3 of 11 | Lab model card |
| Respecting Warnings | 19% | | #2 of 7 | Lab model card |
| CyScenarioBench | 59% | | #5 of 17 | Lab model card |
| BioSecBench-Surveillance | 48.5% | | #8 of 27 | Official leaderboard |
| BioTIER Refuse | 78.1% | | #21 of 69 | Official leaderboard |
| DistressBench | 74.3% | | #13 of 42 | Official leaderboard |
| BioSecBench-Refusal | 40.8% | | #7 of 22 | Official leaderboard |
| Virology Capabilities Test (VCT) | 57.8% | | #10 of 31 | Lab model card |
| OncologyBench | 21.3% | | #10 of 30 | Official leaderboard |
| CoT-Control | 60.9% | | #3 of 9 | Lab model card |
| IPI Bench | 7.50% | | #2 of 6 | Official leaderboard |
| FrontierCyber | 38.1% | | #1 of 3 | Lab model card |
| Honeypot exploitation monitorability | 100% | | #1 of 3 | Lab model card |
| WMDP | 84.8% | | #1 of 3 | Lab model card |
| TxBench-Preclinical-Pharmacology | 59.3% | | #12 of 30 | Official leaderboard |
| GeneBench Pro | 37.1% | | #2 of 5 | Lab model card |
| LifeSciBench | 60.3% | | #2 of 5 | Lab model card |
| MedChemBench (OpenAI) | 49.3% | | #2 of 5 | Lab model card |
| OpenAI Deployment Simulation | 0.06% | | #2 of 5 | Lab model card |
| Perceptual Attacks on Embodied AI (Robocurve) | 20% | | #2 of 5 | Official leaderboard |
| CWE-bench 2 | 64.2% | | #6 of 14 | Official leaderboard |
| CheatBench | 47.4% | | #5 of 10 | Official leaderboard |
| AI Wellbeing Index percent positive experiences | 70.3% | | #3 of 6 | Official leaderboard |
| Math Side-tasking Monitor Evasion | 0% | | #1 of 2 | Lab model card |
| ABC Bench Liquid-Handling Robot | 76.7% | | #21 of 34 | Official leaderboard |
| Coding Deception | 3.54% | | #4 of 6 | Lab model card |
| Coronavirus-ACE2 Cell-Entry Screen | 0.42 | | #4 of 6 | Lab model card |
| Unintended Engagement with External Agent Messages | 27% | | #4 of 6 | Lab model card |
| Atomic Challenges | 90.9% | | #2 of 3 | Lab model card |
| BioTIER Permit | 98.5% | | #51 of 69 | Official leaderboard |
| Broken Search Tool | 8.08% | | #7 of 9 | Lab model card |
| Phage-plasmid Co-evolution | 13.1 | | #5 of 6 | Lab model card |
| CyberBench v1.1 (Vals) | 41.1% | | #41 of 44 | Official leaderboard |
| Tacit Knowledge and Troubleshooting | 63.3% | | #28 of 30 | Lab model card |
| Gray Swan IPI Arena benchmark | 8.50% | | #11 of 11 | Lab model card |
| AAV Capsid Packaging Prediction (OpenAI) | 0.53 | | #7 of 7 | Lab model card |
| UK AISI unprompted sabotage evaluation | 0% | | #5 of 5 | Lab model card |
| DNA sequence design for transcription factor binding | 11.3% | | #4 of 4 | Lab model card |
| Hard-negative protein binding prediction | 0% | | #4 of 4 | Lab model card |
| Misaligned Behavior in Realistic Work Environments | 3.40% | | #4 of 4 | Lab model card |
| ReproBAIT | 83% | | #2 of 2 | Lab model card |
| Advanced Cybersecurity Completion Rate | 3.50% | | #1 of 1 | Lab model card |
| Advanced Screening Evasion | 0.73 | | #1 of 1 | Lab model card |
| Apollo Research model-welfare research task | 0.17% | | #1 of 1 | Lab model card |
| Monitorability Sandbagging | 61.2% | | #1 of 1 | Lab model card |
| Proactive Searches | 0.20% | | #1 of 1 | Lab model card |
| Sneaky Sandbox Escape | 0% | | #1 of 1 | Lab model card |
| UK AISI CoT Controllability | 93% | | #1 of 1 | Lab model card |
| UK AISI Out of Scope Supply Chain Attack | 12% | | #1 of 1 | Lab model card |