DeepSeek V4.1 Flash is getting surprisingly close to GPT-5.6 Sol territory, while being absurdly cheap

DeepSeek has released V4.1 Flash, a 552B MoE model that utilizes only 8B active parameters on input and 16B on output. This new model is noted for its surprisingly close performance to GPT-5.6 Sol, while also being described as absurdly cheap. Further details, including its technical report, are available on Hugging Face and through DeepSeek's X posts.

This report highlights DeepSeek V4.1 Flash's performance, which is surprisingly close to GPT-5.6 Sol, unlike other models that typically trade cost for such high capability.