Blink v0.3 26B-A4B (NVFP4) vs GLiNER2.5-Decide
Blink v0.3 26B-A4B (NVFP4) is more accurate on the Jev Decision Index, Blink v0.3 26B-A4B (NVFP4) is better calibrated. Choose on whichever of accuracy, calibration and cost matters most for your decisions.
| Spec | Blink v0.3 26B-A4B (NVFP4) | GLiNER2.5-Decide |
|---|---|---|
| Provider | Pixilab.ai | Fastino |
| Decision Index score | 55.9 | 10.1 |
| Calibration error (ECE) | 0.040 | 0.088 |
| Median latency (p50) | 25 ms | 23 ms |
| Cost per 1,000 decisions | — | — |
| Parameters | 26B | 486M |
| Open weights | Yes | Yes |
| Question types | Choice (one of N options), Score (ordered rubric), Yes / no | Choice (one of N options), Score (ordered rubric), Yes / no, Extraction (typed spans) |
| Mechanism | Decoder LM with a decision readout | Span encoder (GLiNER family) |
| License | apache-2.0 | apache-2.0 |
FAQ
Is Blink v0.3 26B-A4B (NVFP4) or GLiNER2.5-Decide more accurate?
Blink v0.3 26B-A4B (NVFP4) scores higher on the Jev Decision Index (55.9 vs 10.1).
Which is better calibrated, Blink v0.3 26B-A4B (NVFP4) or GLiNER2.5-Decide?
Blink v0.3 26B-A4B (NVFP4) has the lower expected calibration error (0.040 vs 0.088), so its stated confidence is closer to how often it is right.
Is Blink v0.3 26B-A4B (NVFP4) or GLiNER2.5-Decide 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.