sugiken/Ordis-1.5B-V355-VarGH
TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kPublished:Feb 4, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

The sugiken/Ordis-1.5B-V355-VarGH is a 1.5 billion parameter model, fine-tuned from Qwen2.5-1.5B-Instruct, developed by sugiken. It specializes in practical capabilities such as anti-hallucination, honest refusal, and structured reasoning, making it suitable for applications requiring reliable and truthful responses. This model was trained using a 4-stage progressive training pipeline (PIT) with LoRA, focusing on improving cognitive functions over raw benchmark scores. It demonstrates improved truthfulness on TruthfulQA and strong causal reasoning abilities, despite a small alignment tax on some standard benchmarks.

Loading preview...