Skip to main content

DeepSeek Merges Three Modes: V4.1 Flash Debuts, V4 Pro Retires

DeepSeek Says Goodbye to Mode Switching

Remember when you had to manually switch between DeepSeek's fast mode, expert mode, and image recognition mode? Those days are over. The company announced it has merged all three into a single, unified intelligent mode.

Here's how it works: the model automatically senses how complex your query is. If you upload an image, visual capabilities kick in. If the task is tough, it allocates more processing power. No buttons, no guesswork—just ask and let the AI figure it out.

This move follows the earlier launch of DeepSeek-V4-Flash-Vision-Exp, a multimodal visual understanding model already available on the DeepSeek API platform. Developers can call it via Chat Completions, Messages, or Responses formats.

Image

V4.1 Flash Steals the Show, V4 Pro Heads for the Exit

DeepSeek had already teased the V4.1 Flash release around September 10, 2026 (Beijing Time). After rigorous internal and external testing, the results are in: V4.1 Flash beats V4 Pro on performance, cost, speed, total time—you name it.

So what happens to V4 Pro? It's being retired. Mark your calendar: September 14, 2026, at 12:00 Beijing Time. After that, any V4 Pro requests will be routed to V4.1 Flash and billed at the new rates.

According to a notice in DeepSeek's official communication group, the intermediate version of V4.1 Flash uses a brand-new model structure. It's natively multimodal, more capable, faster, and cheaper to run.

New Prices for the Flash Series

Starting today at 12:00, Flash series pricing gets a shake-up. During off-peak hours, input cache hits cost just 0.02 yuan per unit, cache misses are 1 yuan, and output is 4 yuan. Peak hours? Double those numbers.

For developers and businesses, this could mean significant savings—especially if you can schedule heavy workloads during off-peak windows.

Key Points

  • Unified mode: Fast, expert, and image recognition modes merged into one smart mode that auto-adjusts.
  • V4.1 Flash launch: Outperforms V4 Pro in speed, cost, and capability.
  • V4 Pro retirement: Phased out on September 14, 2026, at 12:00 Beijing Time; requests rerouted to V4.1 Flash.
  • Price update: Flash series pricing adjusted starting September 10, 2026, with off-peak discounts.
  • Multimodal support: V4.1 Flash built on a new natively multimodal architecture.