Data Research and Analysis with Complex Operations (DRACO)
We show this table for reference; we do not rank on it.
Agentic data-analysis tasks scored against per-task rubrics at a 980K-token budget.
Benchmark score on DRACO — November 13, 2026
We compile the DRACO rows from provider self-reports. Claude Opus 5 leads the table at 88.6%, followed by Claude Fable 5.1 (87.7%) and Claude Fable 5.2 (87.7%). We do not use these results to rank models overall.
Claude Opus 5
Anthropic
Reported by AnthropicClaude Fable 5.1
Anthropic
Reported by AnthropicClaude Fable 5.2
Anthropic
Reported by Anthropic10 modelsAgenticCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (10 models)
All 10 results on this page were reported by the labs that built the models.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Opus 5AnthropicReported by Anthropic | 88.6% | Reported by Anthropic | Not reported | Closed |
| 2 | Claude Fable 5.1AnthropicReported by Anthropic | 87.7% | Reported by Anthropic | Not reported | Closed |
| 3 | Claude Fable 5.2AnthropicReported by Anthropic | 87.7% | Reported by Anthropic | Not reported | Closed |
| 4 | Claude Opus 5.5AnthropicReported by Anthropic | 87.4% | Reported by Anthropic | Not reported | Closed |
| 5 | Claude Sonnet 5.5AnthropicReported by Anthropic | 87.0% | Reported by Anthropic | Not reported | Closed |
| 6 | Ling 3.1 FlashInclusionAIReported by InclusionAI | 85.5% | Reported by InclusionAI | Not reported | Open |
| 7 | ERNIE 5.5 Flash PreviewBaiduReported by Baidu | 84.9% | Reported by Baidu | Not reported | Pending |
| 8 | Step 5 PreviewStepFunReported by StepFun | 83.3% | Reported by StepFun | Not reported | Open |
| 9 | Hy4 previewTencentReported by Tencent | 77.2% | Reported by Tencent | Not reported | Open |
| 10 | Ling 3.0 FlashInclusionAIReported by InclusionAI | 70.4% | Reported by InclusionAI | Not reported | Open |
- Lab-reported
- Benchmark-run
- Independent
Among the reported DRACO rows, Claude Opus 5 is first at 88.6%. The third row is 0.9 points behind. The broader top-10 range is 18.2 points, so the table still separates the published systems.
10 models have been evaluated on DRACO. The benchmark falls in the Agentic category. DRACO is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About DRACO
Year
2026
Tasks
Agentic data research and analysis tasks
Format
Normalized rubric score
Difficulty
Professional data analysis
Section 8.10.4 reports the max-effort normalized score using Opus 4.6 as judge and a file-based deliverable protocol. Anthropic warns that the judge change makes the result not directly comparable to the paper headline.
Freshness and provenance
Version
DRACO 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does DRACO measure?
Agentic data-analysis tasks scored against per-task rubrics at a 980K-token budget.
Which model scores highest on DRACO?
Claude Opus 5 by Anthropic currently leads with a score of 88.6% on DRACO.
How many models are evaluated on DRACO?
10 AI models have published results on DRACO in the BenchLM catalog.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.