BankerToolBench - Deliverable - Excel: leaderboard
BankerToolBench - Deliverable - Excel. April2026 primary paper v1,100investment-banking tasks;OpenCode baseline;mean rubric score over three runs;GPT5.2/5.4high reasoning. These are not scores for the October dataset v1.1 update. Deliverable subset=Excel,n=86. Subsets overlap. Metric: Mean rubric score (%). Primary published table; source configurations remain distinct.
Metric: Mean rubric score (%). Source: arxiv.org. Saturation forecast: Around January 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (High) | 57.2 | #78 (GPT-5.4) |
| 2 | GPT-5.2 (High) | 55.5 | #109 (GPT-5.2) |
| 3 | Gemini 3.1 Pro (Preview) | 53.8 | #56 |
| 4 | Claude Opus 4.6 | 51.8 | #60 |
| 5 | Claude Opus 4.5 | 51.3 | #81 |
| 6 | GLM-5 | 45.8 | #136 |
| 7 | Qwen 3.5 397B A17B | 41.6 | #144 |
| 8 | Grok 4 | 30.8 | #176 |
| 9 | Gemini 2.5 Pro | 28.6 | #150 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=bankertoolbench-deliverable-excel · How It Works · Data refreshed daily, snapshot 2026-10-16.