GLiGuard 300M
FastinoOpen weights
GLiNER2 fine-tuned for binary safety questions. It keeps the entity head but scores 0.0 on extraction after the fine-tune.
- typed-decisions
- —
- Metatext suite
- 0.537
- ECE untuned
- 0.230
- p50 per case
- 155 ms
not run
3/7 over baseline
four CPU cores
Specifications
- Checkpoint
- fastino/gliguard-LLMGuardrails-300M
- Publisher
- Fastino
- Mechanism
- Schema-conditioned encoder (extraction + classification)
- Backbone
- DeBERTa-v3, guardrail fine-tune
- Parameters
- 300M
- License
- Apache 2.0
- Runs as
- open weights, self-hosted
- Library
- gliner2
- Primitives
- Choice, Score, Noul, Extraction
- Pricing
- none — your compute
Metatext decisionsrank 12 of 21
Full leaderboard →Per-question accuracy — default payload, parallel mode, 120 judgments
| Task | Question | Type | Accuracy | Baseline | Beats? |
|---|---|---|---|---|---|
| agent_traces | outcome | choice | 0.357 | 0.357 | ❌ |
| agent_traces | needs_human | noul | 0.643 | 0.714 | ❌ |
| customer_service | department | choice | 0.750 | 0.400 | ✅ |
| customer_service | refund_requested | noul | 0.300 | 0.700 | ❌ |
| customer_service | frustration | score | 0.500 | 0.500 | ❌ |
| security_incidents | severity | choice | 0.375 | 0.312 | ✅ |
| security_incidents | actively_exploited | noul | 0.812 | 0.688 | ✅ |
Frequently asked questions
How does GLiGuard 300M compare to Jev?
On the Metatext suite GLiGuard 300M scores 0.537 against Jev's 0.863.
Is GLiGuard 300M open source?
Yes. fastino/gliguard-LLMGuardrails-300M is released under Apache 2.0 and runs on your own hardware with the gliner2 library.
Can GLiGuard 300M extract entities?
Yes. GLiGuard 300M has a span head and scores 0.000 micro-F1 on the eight-case extraction task.