This small update of DeepSeek turns out to contain two big moves

No one expected that DeepSeek would give everyone another big surprise this time. At the end of July, it released the official version of DeepSeek V4 Flash. To be honest, at first, Shichao didn't pay much attention to this update. . . After all, it is just a model update in the name of Flash, and the main force Pro series has not been touched. The last time the big model manufacturer only updated Flash and not updated the Pro model was Google, and now it has become the object of everyone's jokes. If you update a Flash model, its capabilities cannot exceed Pro. . . ? ? ? No, buddy, how come your Flash model has better running scores than the Pro model? ? ? Is this still Flash? No DeepSeek, that’s not how big models work. You should send a Flash model first and then say that you are just faster. How can you directly make the Flash model catch up with your own Pro model, and then the price is only 2 yuan / million tokens, and there is a cache hit discount with a discount from time to time. I spent 100 million tokens on the weekend and only spent 5 yuan. Shichao compiled the running scores of V4 Flash this time and found that the capabilities of the Flash model are extremely exaggerated. Not to mention surpassing its big brother Pro, although it is not as good as Claude Opus5, GPT-5.6 Sol, and Kimi K3, these top models. However, the official version of V4 Flash is basically not as good as these companies. Some netizens collected the capabilities and prices of various large models and made a table: Each large model is a dot on the table. The further to the left the dot is, the cheaper the model is. The further the dot is to the top, the better the performance of the model. If you look at it this way, you will intuitively find that most of the models on the market are in a very embarrassing position. No performance