EXAONE 4.5 lands as open-weight VLM

This is a meaningful multimodal step for EXAONE, but the real story is packaging: long context, multiple runtimes, and a model family tuned for practical deployment rather than just benchmark theater.

• –The model card frames EXAONE 4.5 as a first open-weight VLM from LG AI Research, built on the EXAONE 4.0 stack with a dedicated visual encoder
• –262K context is a standout deployment feature for document-heavy and agentic workflows, especially if the memory footprint is manageable in FP8 or GGUF
• –Benchmark tables position it competitively on document understanding and Korean reasoning, but not as an obvious across-the-board leader
• –The license is non-commercial, which limits adoption for startups and product teams even if the weights are easy to run locally
• –The Reddit reaction suggests the usual LocalLLaMA split: interest in the model family, skepticism about benchmark claims, and annoyance at restrictive licensing