SDR-Bench: leaderboard
SDR-Bench (2026) tests generative sales personalization: given a seller, product, prospect and date, a model with web search restricted to that historical date predicts the pitch points of a later public customer success story. The weighted coverage score is the percent of ground-truth pitch points covered, graded 0 to 5 by a GPT-4o judge over 180 stories; higher is better.
Metric: Weighted coverage score (%) on 180 public customer success stories: a GPT-4o judge grades each ground-truth pitch point 0 to 5 for how well the agent's predicted pitch covers it, normalized to a percentage; agents search a frozen view of the web at the story's historical date; higher is better. Source: arxiv.org. Saturation forecast: Around August 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 55.8 | #161 |
| 2 | GPT-5.4 Mini | 44.63 | #389 |
| 3 | GPT-5.4 | 44.32 | #123 |
| 4 | GPT-4o Mini | 37.46 | #1623 |
| 5 | Qwen 2.5 72B | 36.84 | #1190 |
| 6 | GPT-4o | 35.42 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sdr-bench · How It Works · Data refreshed daily, snapshot 2026-11-13.