AIMultiple - Agentic RAG Routing: leaderboard
AIMultiple Agentic RAG benchmark: each model chooses which of 11 anonymised, deliberately confusable BIRD-SQL databases holds the answer, using per-database schema and query tools (at most 10 API calls, temperature 0). Accuracy of the explicitly declared database on the 184 hardest of 759 questions; %, higher is better, chance 9.1%.
Metric: Routing accuracy (%, 184 hardest of 759 BIRD questions, 11 databases). Source: aimultiple.com. Saturation forecast: Around February 2027. 81 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Argon | 94.02 | #9 |
| 2 | Claude Opus 5.5 | 90.22 | #3 |
| 3 | Muse Wonder | 86.41 | #18 |
| 4 | Gemini 3.7 Flash | 85.33 | #24 |
| 5 | Claude Sonnet 5.5 | 85.33 | #14 |
| 6 | Qwen 3.8 Max (0902) | 85.33 | #36 |
| 7 | Claude Opus 5 | 84.78 | #16 |
| 8 | Kimi K3.1 Pro | 84.78 | #17 |
| 9 | GPT-6 Astra | 84.24 | #6 |
| 10 | Claude Fable 5.1 | 83.7 | #10 |
| 11 | GPT-6.1 Terra | 83.7 | #15 |
| 12 | Kimi K3 | 83.15 | #35 |
| 13 | Qwen 3.8 Max | 83.15 | #34 |
| 14 | Claude Opus 4.8 | 82.61 | #39 |
| 15 | MiMo-V2.6-Pro | 82.61 | #48 |
Interactive version: theaggregate.ai/benchmark?slug=aimultiple-agentic-rag-routing · How It Works · Data refreshed daily, snapshot 2026-11-05.