DRACO: leaderboard
Perplexity DRACO deep-research agent evaluation reported in Anthropic's Claude Opus 4.8 system card.
Metric: Score (%). Source: arxiv.org. Source: leaderboard.steel.dev. Saturation forecast: Around January 2027. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 88.6 | #12 |
| 2 | Claude Opus 5.5 | 87.4 | #3 |
| 3 | Claude Sonnet 5.5 | 87 | #11 |
| 4 | Claude Mythos 5 | 86.4 | #2 |
| 5 | Claude Mythos Preview | 83.6 | #4 |
| 6 | Claude Haiku 5.5 | 81.5 | #40 |
| 7 | Claude Opus 4.8 | 80.4 | #37 |
| 8 | Claude Opus 4.7 | 77.7 | #49 |
| 9 | MiniMax-M3 | 73.23 | #136 |
| 10 | Claude Fable 5 | 65.3 | #16 |
| 11 | Claude Opus 4.6 | 59.8 | #63 |
| 12 | O3 Deep Research | 52.1 | #115 |
| 13 | Claude Opus 4.5 | 46.7 | #83 |
| 14 | O4 Mini Deep Research | 41.9 | #224 |
No result here: #6 GPT-6 Astra, #7 GPT-6.1 Sol, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=draco · How It Works · Data refreshed daily, snapshot 2026-10-20.