SalesBench: leaderboard
Long-horizon agent-to-agent sales simulation: models work a 50-lead list selling to LLM buyer personas over multi-turn calls; scored by realized reward.
Metric: 50-Lead Reward. Source: hamzamostafa.com. Saturation forecast: Around August 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.1 | 0.51 | #97 |
| 2 | Claude Opus 4.7 | 0.5 | #49 |
| 3 | GPT-5.5 | 0.32 | #31 |
| 4 | Kimi K2.6 | 0.27 | #104 |
| 5 | Qwen 3.6 Max Preview | 0.17 | #85 |
| 6 | DeepSeek V4 Pro | 0.11 | #102 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=salesbench · How It Works · Data refreshed daily, snapshot 2026-10-20.