#1 on S1MB english-v1: Borda 90.38 across 137 benchmarks (2026-10-20)
Darwin-27B-ZTC-v9 is #1 on the S1MB English v1 leaderboard, read from the live board on 20 October 2026.
| Rank | Model | Borda |
|---|---|---|
| 1 | Darwin-27B-ZTC-v9 | 90.38 |
| 2 | kettei-fat v1.0 | 90.14 |
| 3 | Superion S1 | 89.53 |
| 4 | D3 (27B) | 89.28 |
| 5 | Darwin-27B-ZTC-v5 | 88.44 |
197 models, 137 benchmarks.
On the ranking method. S1MB ranks by Borda count: each benchmark is ordered separately and the ranks are aggregated. A few strong tasks will not carry you, because every benchmark contributes the same weight regardless of how its metric happens to be scaled. The board also publishes a task-average column, where this model is second at 72.20. Borda is the board's default ranking and the headline above refers to it.
What the model does. Given a situation and a typed question, it returns a calibrated probability for every option in a single forward pass and generates no text. Three question shapes: yes or no, pick one, and a score on a scale. Decision benchmarks score calibration, not only accuracy, which is what lets a team set an action threshold on the number.
Serving. Standard /v1/systemone endpoint of transformers serve:
hf download FINAL-Bench/Darwin-27B-ZTC-v9 systemone_config.json --local-dir .
transformers serve FINAL-Bench/Darwin-27B-ZTC-v9 --decision-config systemone_config.json
Weights are Apache-2.0. Public boards move as new entries arrive, so check the date on any position quoted here.
Leaderboard: https://huggingface.co/spaces/hotchpotch/S1MB-leaderboard