TopBench (Agentic) - Decision Making: leaderboard
TopBench (2026) tests implicit predictive reasoning in table question answering: the answer is not in the table and must be predicted from historical rows. Share of decision questions whose chosen option matches the reference, from 0 to 1, solved by an agent that runs Python in a sandbox; higher is better.
Metric: Decision accuracy (0 to 1) on TopBench's 186 decision-making questions, which ask which of several described profiles will have the best predicted outcome given a historical table, solved by the authors' ReAct agent that writes and runs Python in a Docker sandbox over the table: share of questions whose final choice matches the reference; values and decisions extracted from the free-text answer by a DeepSeek-V3.2 judge and checked against the answer text; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 0.65 | #175 |
| 2 | DeepSeek V3.2 | 0.59 | #419 |
| 3 | Qwen 3 235B A22B 2507 (Thinking) | 0.57 | #466 |
| 4 | Claude Sonnet 4.5 | 0.56 | #261 |
| 5 | GPT-5.2 | 0.55 | #185 |
| 6 | Qwen 3 235B A22B 2507 Instruct | 0.45 | #616 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=topbench-agentic-decision-making · How It Works · Data refreshed daily, snapshot 2026-11-21.