Jev directory
CommunityJev Models & Reimplementations
GLiNER2.5-Decide-1B
Fastino's 1B English member of the GLiNER2.5 classification family: pass any label set at call time for intent, routing, sentiment or multi-label tags and get labels back in a single forward pass with no generated tokens. It averages 59.6% exact-match on Fast Decisions.
- Reading time
- 2 min
- Published by
- Community
- Author
- Fastino
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- 1B-parameter English classifier built on fastino/gliner2-xl-0111; labels are passed at call time, not fixed.
- One forward pass scores the labels with no prompt template and no generated tokens; several heads can share a call.
- Averages 59.6% exact-match on the 17-domain Fast Decisions suite, just under the 340M GLiNER2.5-Decide at 60.2%.
- Load through AutoExtractor.from_pretrained("fastino/GLiNER2.5-Decide-1B") after pip install gliner2.
- Apache-2.0; the card positions it for the same operational decisions as the rest of the family.
Quickstart
python
from gliner2 import AutoExtractor
model = AutoExtractor.from_pretrained("fastino/GLiNER2.5-Decide-1B")Watch out
Card reviewed 2026-10-17; the 59.6% average is Fastino's self-reported exact-match result on its own Fast Decisions suite, and the card notes the suite is English. The card states the model does not reason, explain or answer open questions, and no independent rerun is linked.
Questions about this entry
Short answers drawn from the linked project or page.
The 1B English classification model in Fastino's GLiNER2.5 family. It takes text plus a label set supplied at call time — intent, routing, sentiment, priority, policy, multi-label tags — and scores them in a single forward pass with no generated tokens.
Size and slightly different scores: the 1B averages 59.6% exact-match on Fast Decisions against 60.2% for the 340M model, and it is built on a different base, fastino/gliner2-xl-0111, versus fastino/gliner2-large-v1. Both are English and Apache-2.0.
A local Python interface: pip install gliner2, then AutoExtractor.from_pretrained("fastino/GLiNER2.5-Decide-1B") and classify_text with named heads. The card does not document a System One endpoint, but the pattern matches Jev: score supplied labels instead of generating text.
Exact-match accuracy on fastino/fast-decisions, 17 domains with 300 held-out examples each and identical text and labels for every model in the comparison table: 60.2% GLiNER2.5-Decide, 59.6% this model, 57.6% JevK5, 56.7% multi, 56.4% SemIf, 49.0% GLiFormer and 46.6% Laya Router.
Single-label tasks return one string, such as an intent name; multi-label tasks return every label above the model's threshold. A single call can score several heads at once, so one message can be classified for intent, sentiment and priority together.
Fastino, as part of the GLiNER2.5 release line. The card links the GLiNER2 paper (arXiv 2507.18546) and the fastino-ai/GLiNER2 GitHub repository, and the weights are Apache-2.0.
No. The card says it does not reason, explain or answer open questions; it is a specialist for operational decisions such as customer and banking intent, review sentiment, document type, ticket routing, handoff, moderation, severity, urgency and spam.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Standard One SH (StandardThinking)
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
Hacker News: Decision Index 0.2.1 scores open reproductions of Jev on a frozen suite.
jackbravo4 pointsDecision Index 0.2.1 – 70 open reproductions of TypeSafe Jev's Decision ModelRead the thread on Hacker NewsHacker News: Pointer to a DiffusionGemma plus vLLM tutorial for a Jev-shaped classifier.
1 pointHow to build a Jev-style classifier with DiffusionGemma and vLLMRead the thread on Hacker NewsMore Jev resources like this
Fastino's 340M English classification model and the headline member of the GLiNER2.5 family: pass any label set at call time and get intent, routing, sentiment or multi-label tags back in one forward pass, at 60.2% exact-match average on the Fast Decisions suite.
Jev Models & Reimplementations#community#open-models#local
Community
The 322M mmBERT-base build of Laya for 100+ languages: state plus typed questions return typed answers with probabilities in one forward pass, with up to 8,192 tokens for long documents.
Jev Models & Reimplementations#community#open-models#local
Community
The Hugging Face hub for Laya's open System 1 checkpoints: the 421M English model at the repo root, a 322M multilingual build in a subfolder, and the typed-decisions fine-tune, all Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
A DistilBERT prototype from akumar0205 that sorts historical email into legitimate and spam, reporting 0.987 accuracy and 0.983 spam F1 on a held-out SpamAssassin split; it is not a phishing or malware detector.
Jev Models & Reimplementations#community#open-models#local
Community
Fastino's 287M multilingual classification model in the GLiNER2.5 family: pass any label set at call time and get labels back in a single forward pass, with no prompt template and no generated tokens. It averages 56.7% exact-match on the Fast Decisions suite.
Jev Models & Reimplementations#community#open-models#local
Community
TokenRhythm's 4B prefill-only decision model built on NeoHorse-1-4B: given a state and application-defined questions it returns Choice, Noul, and Score probabilities without generating text, reporting a 77.70 six-group aggregate under Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.