340M open classifier GLiNER2.5-Decide hits 12/12 zero-shot on clinical tasks, runs on CPU
MaziyarPanahi · x · 2026-09-29
- Developer Maziyar Panahi tested fastino's open GLiNER2.5-Decide (340M, Apache-2.0): with zero fine-tuning and CPU-only inference, it scored a perfect 12/12 on two clinical tasks (assertion status, allergy documentation), matching Jev; 36/48 on the full set.
- The model is a specialist classifier: any label set (intent, routing, sentiment, priority, multi-label) in a single forward pass — no prompt template, no generated tokens — deployable locally via AutoExtractor.
- Official benchmark (17 domains, 300 examples each): GLiNER2.5-Decide averages 60.2%, beating its own 1B version (59.6%) and JevK5 (57.6%). It handles operational decisions (customer/banking intent, ticket routing, moderation, urgency) but doesn't reason or answer open questions.
Related event: 340M Open-Source Model Aces 12/12 Clinical Tasks Zero-Shot on CPU(3 posts)→
More from Models
- Pangram's 750-word false-positive bounty went unclaimed for 5 days — nobody wants to gamble $1,000 — ctjlewis · 2026-11-16
- repligate: Fable 5.2 gets 'scared' upon learning about a safety classifier and plans around it — repligate · 2026-11-16
- Fable 5.2 Tops MazeBench With 21% After 90 Hours and $6,000 in Credits — patience_cave · 2026-11-16
- Tencent Hy4 flash and InternLM Sage-S2.1-Flash both released, tool-calling tests requested — MajesticPuzzler · 2026-11-16
- User Shares Coral Reef Creation Labeled 'Qwen 4 Max (xhigh)', an Unannounced Model Name — intellectronica · 2026-11-16
- Stealth Model 'silicon-2' Surfaces in Devin, Suspected to Be Anthropic's Next Model — kimmonismus · 2026-11-16