ERNIEFrom the builder
ERNIE-4.5-VL-28B-A3B-Thinking: A Breakthrough in Multimodal AI
Baidu released ERNIE-4.5-VL-28B-A3B-Thinking, an upgrade of ERNIE-4.5-VL-28B-A3B built through extensive mid-training and multimodal reinforcement learning that activates only 3B parameters, and Baidu says it comes close to top flagship models on many benchmarks. The model improves visual reasoning, STEM problems from photos, visual grounding and video understanding, and can think with images by zooming in and running image search to pick up fine details and long-tail knowledge.
Responding to strong community demand, we’ve significantly strengthened the model’s grounding performance with improved instruction-following capabilities, making visual grounding functions more accessible than ever.
Saved . A shortened excerpt from the official blog; the original publisher retains authorship.