Fastino releases GLiNER2.5-Decide: a 340M open-weight decision model that runs on CPU

Fastino Labs has released GLiNER2.5-Decide, a 340M-parameter open-weight non-generative model for judgment calls in agent pipelines: routing, triage, tool selection and guardrails. Weights ship under Apache 2.0 and run on CPU, GPU or air-gapped environments via pip install gliner2.

• DeBERTa-v3-large based model returns answers with probability distributions and feasibility metadata
• Joint decoding enforces rules across answers, e.g. harm_type and safety=unsafe together
• Leads 9 of 17 Fast Decisions datasets, scoring 75.3% on support intent and 64.3% on banking intent
• p50 latency: 167.3 ms on Xeon CPU, 38.3 ms on V100 for short inputs