00:00
2026-07-24
rocm.blogs.amd.com
large-language-models
Enabling Language-specific Reasoning in Multilingual Models with Reinforcement Learning
The Poro 2 Long family of models, developed by LumiOpen in collaboration with TurkuNLP and the OpenEuroLLM project, extends the context window of the Poro 2 base model to 128k tokens while maintaining…