DeepSeek has released two distinct versions of its V4 model: DeepSeek V4 Pro and DeepSeek V4 Flash. The V4 Pro is a larger model with 1.6 trillion total parameters and 49 billion active parameters, designed for complex tasks like reasoning, coding, and research. In contrast, the V4 Flash is an efficiency-focused model with 284 billion total parameters and 13 billion active parameters, suitable for high-volume, latency-sensitive, and cost-controlled workloads. Both models share a 1 million token context window and support advanced features like thinking and non-thinking modes, JSON output, and tool calls, but differ significantly in pricing and architecture. AI
IMPACT Offers developers distinct choices for different workloads, balancing capability with cost and latency.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
