Iteration on Meta's proprietary Superintelligence Labs model, with gains in reasoning and coding.
| Benchmark | Score | Rank |
|---|---|---|
aaAgentic | 37.5% | #4 / 5 |
hleArtificial Analysis | 45.1% | #5 / 66 |
AA Intelligence Independent composite of agentic work, coding, scientific reasoning, and general capability | 50.6% | #9 / 11 |
AA Coding Independent composite of implementation, terminal use, and real-codebase agent performance | 71.3% | #9 / 11 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 88.7% | #13 / 51 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 86.6% | #13 / 51 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 85.9% | #18 / 53 |
GPQAArtificial Analysis PhD-level science questions even experts struggle with | 89.8% | #25 / 78 |