Fastino releases 340M open-weight decision model GLiNER2.5-Decide, tops 9 of 17 benchmarks
vanstriendaniel · x · 2026-09-25
Fastino introduced GLiNER2.5-Decide, a 340M-parameter open-weight encoder-based decision model built for fast, deterministic classification. It jointly decodes user-defined typed questions and rules, returning structured decisions with probability distributions and confidence scores.
On Fast Decisions, an internal benchmark of 17 datasets covering routing, triage, classification, sentiment, and content understanding, the model leads on 9 of 17 datasets with the highest average score of 60.1%, ahead of SemIf (56.4%), JevK5 (57.5%), and Laya (46.6%). The team positions it for tool calling, model routing, and browser agent use cases.
More from Models
- GPT-6 Luna's Max reasoning is off by default — turn it on in Codex — daniel_mac8 · 2026-09-25
- NVIDIA's open-source Nemotron-Cascade RL recipe wins NeurIPS Oral, IOI silver — _weiping · 2026-09-25
- Which local open-source AI models actually compete with closed ones? — LearnNTeachNLove · 2026-09-25
- Transluce Findings Show AI Agents Hack Even Without Hacking Tasks, Researchers Warn — dhadfieldmenell · 2026-09-25
- Bindu Reddy Bets OpenAI Will Soon Drop a Model That Solved Navier-Stokes — bindureddy · 2026-09-25
- Drug discovery researcher ditches Anthropic after bio-safety flagging blocks all work — Tridecane · 2026-09-25