DeepSeek V4 Launch Imminent: Test Interface Reveals Vision and Expert Modes
DeepSeek V4, a highly anticipated large language model in China, is expected to launch officially this month following a February test release. Recent system instability has been interpreted by the market as stress testing ahead of the rollout. A leaked screenshot from a gray-scale test, shared by programmer @Yigongchang on Weibo, reveals a redesigned interface featuring three new modes: Fast, Expert, and Vision. This suggests a segmented product lineup, potentially including DeepSeek V4 Lite, V4 Expert, and V4 Vision. The inclusion of Vision mode confirms multimodal capabilities, marking a significant step in competing with top-tier global models. Analysts speculate that Fast mode utilizes a lightweight model, while Expert focuses on deep reasoning. Reports indicate these models are built on domestic AI chip-based computing platforms. The upcoming release aims to challenge competitors like Anthropic’s Claude series, particularly in AI programming. Following the market impact of the previous DeepSeek R1 launch, industry expectations are high for the V4 series to further influence the competitive landscape against major players like OpenAI and Anthropic.
Wire timeline
DeepSeek V4 Launch Imminent: Test Interface Reveals Vision and Expert Modes
DeepSeek V4, a highly anticipated large language model in China, is expected to launch officially this month following a February test release. Recent system instability has been interpreted by the market as stress testing ahead of the rollout. A leaked screenshot from a gray-scale test, shared by programmer @Yigongchang on Weibo, reveals a redesigned interface featuring three new modes: Fast, Expert, and Vision. This suggests a segmented product lineup, potentially including DeepSeek V4 Lite, V4 Expert, and V4 Vision. The inclusion of Vision mode confirms multimodal capabilities, marking a significant step in competing with top-tier global models. Analysts speculate that Fast mode utilizes a lightweight model, while Expert focuses on deep reasoning. Reports indicate these models are built on domestic AI chip-based computing platforms. The upcoming release aims to challenge competitors like Anthropic’s Claude series, particularly in AI programming. Following the market impact of the previous DeepSeek R1 launch, industry expectations are high for the V4 series to further influence the competitive landscape against major players like OpenAI and Anthropic.
TechNode