Flash-tier successor to 3.6 posting GPQA 94.5 and ARC-AGI-2 85.3 — frontier-level reasoning at the cost-efficient tier.
| Benchmark | Score | Rank |
|---|---|---|
GPQAArtificial Analysis PhD-level science questions even experts struggle with | 94.5% | #2 / 88 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 88.7% | #3 / 59 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 89% | #3 / 57 |
hleArtificial Analysis | 47.9% | #3 / 76 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 90.1% | #5 / 58 |
ARC-AGIARC Prize Novel reasoning tasks requiring fluid intelligence | 85.3% | #5 / 30 |
AA Coding Independent composite of implementation, terminal use, and real-codebase agent performance | 76.1% | #8 / 18 |
aaAgentic | 45.1% | #9 / 12 |
AA Intelligence Independent composite of agentic work, coding, scientific reasoning, and general capability | 56% | #10 / 18 |