GLiNER2.5-Decide vs Winnow-12B
Winnow-12B is more accurate on the Jev Decision Index, GLiNER2.5-Decide is better calibrated. Choose on whichever of accuracy, calibration and cost matters most for your decisions.
| Spec | GLiNER2.5-Decide | Winnow-12B |
|---|---|---|
| Provider | Fastino | EldanRing |
| Decision Index score | 10.1 | 46.6 |
| Calibration error (ECE) | 0.088 | 0.168 |
| Median latency (p50) | 23 ms | 73 ms |
| Cost per 1,000 decisions | — | — |
| Parameters | 486M | 12B |
| Open weights | Yes | Yes |
| Question types | Choice (one of N options), Score (ordered rubric), Yes / no, Extraction (typed spans) | Choice (one of N options), Score (ordered rubric), Yes / no |
| Mechanism | Span encoder (GLiNER family) | Decoder LM with a decision readout |
| License | apache-2.0 | apache-2.0 |
FAQ
Is GLiNER2.5-Decide or Winnow-12B more accurate?
Winnow-12B scores higher on the Jev Decision Index (46.6 vs 10.1).
Which is better calibrated, GLiNER2.5-Decide or Winnow-12B?
GLiNER2.5-Decide has the lower expected calibration error (0.088 vs 0.168), so its stated confidence is closer to how often it is right.
Is GLiNER2.5-Decide or Winnow-12B cheaper per 1,000 decisions?
A hosted price is not recorded for both models; open-weight models cost the compute you run them on.
Cost per 1,000 decisions assumes 200 input tokens per decision. Pricing is indicative — confirm with the provider before production use. Highlighted values indicate the better figure for that row.