← All benchmarks
DecisionPivot core accuracy
General QA · 2026-10-20
Item accuracy on DecisionPivot's core track of 7,800 generated items in 13 task families, macro-averaged over families. Diagnostic, because the protocol differs by model.
Benchmark source: arXiv
Top models (higher is better)
| Model | Score | Source |
|---|
| GLM-5.3 | 91.7 | DecisionPivot paper Table D10 (GLM-5.3, lowest reasoning effort, accuracy) · StartLux |
| Gemma 4 31B IT | 67.5 | DecisionPivot paper Table D10 (Gemma-4-31B-it, accuracy) · StartLux |
| Gemma 4 26B A4B IT | 56.6 | DecisionPivot paper Table D10 (Gemma-4-26B-A4B-it, accuracy) · StartLux |
| Qwen3.8 27B | 53.6 | DecisionPivot paper Table D10 (Qwen3.8-27B, accuracy) · StartLux |
| Qwen3.6 35B A3B | 49.1 | DecisionPivot paper Table D10 (Qwen3.6-35B-A3B, accuracy) · StartLux |
| Qwen3.5-9B | 47.9 | DecisionPivot paper Table D10 (Qwen3.5-9B, accuracy) · StartLux |
| Qwen3.5-4B | 46.6 | DecisionPivot paper Table D10 (Qwen3.5-4B, accuracy) · StartLux |
| Phi-4 | 46.5 | DecisionPivot paper Table D10 (Phi-4, accuracy) · StartLux |
| Gemma 3 12B IT | 44.7 | DecisionPivot paper Table D10 (Gemma-3-12B-it, accuracy) · StartLux |
| Granite 3.3 8B Instruct | 42.0 | DecisionPivot paper Table D10 (Granite-3.3-8B-Instruct, accuracy) · StartLux |
| Qwen3.5-2B | 40.8 | DecisionPivot paper Table D10 (Qwen3.5-2B, accuracy) · StartLux |
| Qwen3.5-0.8B | 40.0 | DecisionPivot paper Table D10 (Qwen3.5-0.8B, accuracy) · StartLux |