Knowledgator Releases GLiFormer: A 575M-Parameter Encoder That Hits 91.10 F1 on Nested JSON Extraction Without Generating Tokens

Knowledgator Engineering introduced GLiFormer, a schema-conditioned encoder for information extraction combining NER, classification, relation extraction, nested JSON, and embeddings in one model. The Large version with 575.6 million parameters achieves 91.10 F1 on nested JSON, with Apache 2.0 weights available on Hugging Face.

• GLiFormer Base: 264.2M parameters, Large: 575.6M, both Apache 2.0
• Large scores 91.10 F1 on nested JSON vs 91.96 for GPT-5.6-luna
• Median latency for Base: 69 ms on RTX PRO 6000 GPU and 547 ms on CPU
• Model does not generate tokens: values taken as spans from source text