DeepSeek-V4-Flash Official Release: 13B Activated Parameters Fuel Agent Market
DeepSeek-V4-Flash API officially launched its public beta on July 31, signaling a major strategic move for the domestic AI leader into the Agent ecosystem. Breaking from its previous "Pro-strong, Flash-weak" dynamic, the updated Flash version now matches or exceeds the performance of the V4-Pro preview from three months ago across multiple Agent benchmarks.
According to official metrics, V4-Flash achieved scores of 82.7 on Terminal Bench 2.1, 76.7 on Cybergym, and 70.3 on Toolathlon verified. With 284 billion total parameters but only 13 billion activated, this model uses nearly one-tenth the compute of V4-Pro (49 billion activated parameters). DeepSeek confirmed that this update relied solely on re-training without structural changes, demonstrating that optimized training and data quality can significantly narrow the performance gap between lightweight and flagship models.

Another key feature is the debut of DeepSeek Harness, the company’s proprietary Agent framework. V4-Flash natively supports OpenAI’s Responses API format and includes specific optimizations for Codex, enabling developers to migrate OpenAI-based Agent applications to DeepSeek with minimal cost and effort.
The strategy is clear: capture the Agent market using a high-performance, cost-effective lightweight model. Since Agent workflows are highly sensitive to inference speed and pricing, Flash’s affordability provides a distinct advantage in multi-turn tasks requiring frequent tool calls.
Coming less than two months after DeepSeek’s first funding round exceeding 50 billion yuan, the company’s valuation now stands at approximately $71 billion. Backed by this capital, DeepSeek is transitioning from its previous "no funding, no listing" stance to an aggressive commercialization phase. With the official V4-Pro release imminent, the competition in the Agent space has officially begun.
Related article
Swedish AI Startup Lovable Eyes $13.2 Billion Valuation After Major Funding Round
As AI-driven coding tools gain traction, Swedish startup Lovable has secured a major funding round. The company aims to raise $3 billion, potentially boosting its valuation to $13.2 billion—double the $6.6 billion recorded last December. Menlo Ventur
Google Tests Remy AI Agent for Gemini as Focus Shifts to User Control
According to Business Insider, Google is testing Remy, a new AI personal agent for Gemini. This tool aims to execute tasks on behalf of users, streamlining both professional workflows and daily routines.Currently, Remy is undergoing testing in an int
How to fix Core Web Vitals for better SEO rankings
Streamline Report Card Comments with AI ToolsIntroductionAI Tools for Generating Report Card CommentsMagic SchoolAlmanac AIChat GPTUsing Magic School to Generate Report Card CommentsLogging into Magic SchoolSelecting the Report Card Comments ToolCust
Related Special Topic Recommendations
Comments (0)
0/500
DeepSeek-V4-Flash API officially launched its public beta on July 31, signaling a major strategic move for the domestic AI leader into the Agent ecosystem. Breaking from its previous "Pro-strong, Flash-weak" dynamic, the updated Flash version now matches or exceeds the performance of the V4-Pro preview from three months ago across multiple Agent benchmarks.
According to official metrics, V4-Flash achieved scores of 82.7 on Terminal Bench 2.1, 76.7 on Cybergym, and 70.3 on Toolathlon verified. With 284 billion total parameters but only 13 billion activated, this model uses nearly one-tenth the compute of V4-Pro (49 billion activated parameters). DeepSeek confirmed that this update relied solely on re-training without structural changes, demonstrating that optimized training and data quality can significantly narrow the performance gap between lightweight and flagship models.

Another key feature is the debut of DeepSeek Harness, the company’s proprietary Agent framework. V4-Flash natively supports OpenAI’s Responses API format and includes specific optimizations for Codex, enabling developers to migrate OpenAI-based Agent applications to DeepSeek with minimal cost and effort.
The strategy is clear: capture the Agent market using a high-performance, cost-effective lightweight model. Since Agent workflows are highly sensitive to inference speed and pricing, Flash’s affordability provides a distinct advantage in multi-turn tasks requiring frequent tool calls.
Coming less than two months after DeepSeek’s first funding round exceeding 50 billion yuan, the company’s valuation now stands at approximately $71 billion. Backed by this capital, DeepSeek is transitioning from its previous "no funding, no listing" stance to an aggressive commercialization phase. With the official V4-Pro release imminent, the competition in the Agent space has officially begun.
Swedish AI Startup Lovable Eyes $13.2 Billion Valuation After Major Funding Round
As AI-driven coding tools gain traction, Swedish startup Lovable has secured a major funding round. The company aims to raise $3 billion, potentially boosting its valuation to $13.2 billion—double the $6.6 billion recorded last December. Menlo Ventur
How to fix Core Web Vitals for better SEO rankings
Streamline Report Card Comments with AI ToolsIntroductionAI Tools for Generating Report Card CommentsMagic SchoolAlmanac AIChat GPTUsing Magic School to Generate Report Card CommentsLogging into Magic SchoolSelecting the Report Card Comments ToolCust





Home






