OpenDFM/RetroDFM-R-8B
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Jan 12, 2026License:gpl-3.0Architecture:Transformer0.0K Open Weights Cold

OpenDFM/RetroDFM-R-8B is an 8 billion parameter reasoning-driven large language model developed by OpenDFM, specifically designed for chemical retrosynthesis prediction. Unlike traditional models, it integrates large-scale reinforcement learning with chemically verifiable rewards, leading to enhanced generalization, reliability, and interpretability. This model excels at reconstructing multistep chemical routes and provides explicit, human-interpretable reasoning for its predictions, outperforming existing state-of-the-art approaches on standard benchmarks.

Loading preview...