GLiNER 2.5 multi
FastinoOpen weights
The multilingual 2.5-generation GLiNER2 checkpoint (4,096-token context).
- typed-decisions
- —
- Metatext suite
- 0.494
- ECE untuned
- 0.244
- p50 per case
- 346 ms
not run
4/7 over baseline
four CPU cores
Specifications
- Checkpoint
- fastino/gliner2.5-multi-v1
- Publisher
- Fastino
- Mechanism
- Schema-conditioned encoder (extraction + classification)
- Backbone
- mDeBERTa-v3 (boundary pairing)
- Parameters
- 287M
- License
- Apache 2.0
- Runs as
- open weights, self-hosted
- Library
- gliner2
- Primitives
- Choice, Score, Noul, Extraction
- Pricing
- none — your compute
Metatext decisionsrank 15 of 21
Full leaderboard →Per-question accuracy — default payload, serial mode, 120 judgments
| Task | Question | Type | Accuracy | Baseline | Beats? |
|---|---|---|---|---|---|
| agent_traces | outcome | choice | 0.286 | 0.357 | ❌ |
| agent_traces | needs_human | noul | 0.357 | 0.714 | ❌ |
| customer_service | department | choice | 0.600 | 0.400 | ✅ |
| customer_service | refund_requested | noul | 0.550 | 0.700 | ❌ |
| customer_service | frustration | score | 0.650 | 0.500 | ✅ |
| security_incidents | severity | choice | 0.375 | 0.312 | ✅ |
| security_incidents | actively_exploited | noul | 0.750 | 0.688 | ✅ |
Frequently asked questions
How does GLiNER 2.5 multi compare to Jev?
On the Metatext suite GLiNER 2.5 multi scores 0.494 against Jev's 0.863.
Is GLiNER 2.5 multi open source?
Yes. fastino/gliner2.5-multi-v1 is released under Apache 2.0 and runs on your own hardware with the gliner2 library.
Can GLiNER 2.5 multi extract entities?
Yes. GLiNER 2.5 multi has a span head and scores 0.885 micro-F1 on the eight-case extraction task.