DeepSeek releases DeepSeek V3.2
Another fine-tuning iteration on the V3 base, focused on lowering peak memory use and speeding up inference on local hardware.
MODEL
English reports linked to this entity will appear here as their translations are published.