DeepSeek releases V4.1 Flash model
已归档 —— 这条已轮出今日牌堆，完整内容在此保留。
一句话看懂
DeepSeek released V4.1-Flash on September 10, 2026, its smallest new-architecture model.
The interim model adds native multimodal support as DeepSeek prepares a STAR Market IPO.
背景
DeepSeek has followed a rapid 2026 release cadence, shipping point updates to its V4 family every few weeks with build tags tied to release dates. V4.1-Flash is positioned as an interim model between the current V4 Flash and V4 Pro lines, built on a brand-new architecture that makes multimodal capabilities native rather than bolted on. DeepSeek said the model offers stronger capability, faster inference, higher throughput and lower cost, though it has not presented it as a formal release and no technical report or benchmark scores have been published.
来龙去脉
- Sep 8, 2026DeepSeek quietly opened a limited beta for V4.1 Flash under the model ID deepseek-v4.1-flash-expires-on-0910, with a two-day testing window and pricing matching V4 Flash.
- Sep 9, 2026Chinese tech outlets including IT Home, Sina Finance and Ifeng Tech reported a DeepSeek notice that V4.1 Flash was scheduled for release around September 10, Beijing time.
- Sep 10, 2026DeepSeek released V4.1-Flash, described as the smallest model in its new architecture family, alongside a Flash price cut and a temporary routing change sending V4 Pro API requests to the new model.
各方怎么说
- Official
- DeepSeek stated that V4.1 Flash adopts a new model architecture with native multimodal support, stronger capabilities, faster speed and lower cost.
- Analysts
- Independent benchmark sites have not yet published scores for the new model, so coverage is limited to confirmed details until test results appear.
待核实
Independent benchmark scores for V4.1 Flash have not been published. The claim that DeepSeek is preparing an initial public offering on Shanghai's STAR Market comes from a Reuters report cited by Street Insider and has not been independently confirmed in the retrieved material.