Open-source and open-weight models developed by Fastino Labs.
Our models
9 MODELS · 45M+ DOWNLOADS
NEW
SEPTEMBER 24, 2026 · DECISION
An open weight 355M parameter model that returns typed decisions with confidence scores. GLiNER2.5-Decide can also return spans and relations, enforce constraints between labels, and support both full and LoRA fine-tuning. It runs locally on CPUs and can be deployed in air-gapped environments.
NEW
AUGUST 24, 2026 · NER
Latest GLiNER model for zero-shot entity extraction and classification. GLiNER2.5 replaces span enumeration with boundary prediction, adding unlimited span length, joint entity-relation extraction, constrained classification, span attributes, and long-context extraction.
AUGUST 11, 2026 · FINANCE
A 30B-parameter mixture-of-experts model with 3B active parameters specialized model for finance domain, fine-tuned on Nemotron 3.5 Lightning with Fastino Fine-Tuning Agent.
AUGUST 11, 2026 · HEALTHCARE
A 30B-parameter mixture-of-experts model with 3B active parameters specialized model for healthcare domain, fine-tuned on Nemotron 3.5 Lightning with Fastino Fine-Tuning Agent.
JULY 8, 2026 · PRIVACY, PII, & SAFETY
A 300M-parameter small language model that covers both PII detection and safety moderation in a single forward pass.
MAY 13, 2026 · PRIVACY & PII
A 300M-parameter multilingual model that detects and redacts personally identifiable information across 42 entity types, with a schema you can customize at inference time.
MAY 2026 · SAFETY
Lightweight guardrail model for prompt injection attempts, jailbreaks, and unsafe content classification. Designed to sit inline with agent loops at sub-50ms latency.
MAR 2025 · INFORMATION EXTRACTION
A 205M-parameter small language model for zero-shot entity extraction, text classification, and structured parsing. The second-generation model in the GLiNER family, with a single encoder handling NER, classification, and JSON output.
FEB 2024 · OPEN-VOCABULARY NER
The original open-vocabulary NER model in GLiNER family.
Fine-tune open weight models with the Fastino Fine-Tuning Agent
Deploy production-grade models that you own in hours, not weeks.
Get access