DeepSeek's preview release of the V4 model family on April 24, 2026—including the 1.6T-parameter V4-Pro (49B active) and lighter V4-Flash (284B, 13B active)—marks a significant escalation in open-source AI competition, with both supporting 1M-token context and agentic optimizations via novel hybrid attention and domestic chip support like Huawei Ascend. The models debuted swiftly on the LMSYS Chatbot Arena text leaderboard, where V4-Pro (thinking mode) holds #20 overall (Elo 1463, preliminary) and #2 among open models, matching Kimi-K2.6 while leading in medical/healthcare; V4-Flash ranks #47-#55. This reflects trader consensus on DeepSeek's benchmark dominance in math, coding, and STEM, challenging U.S. frontier models like GPT-5.4, though sustained Elo gains depend on accumulating user votes amid intensifying Chinese lab rivalries from GLM and Moonshot. Full V4 rollout and Arena maturation remain key catalysts.
Experimental AI-generated summary referencing Polymarket data. This is not trading advice and plays no role in how this market resolves. · UpdatedView resolved

Beware of external links.
Beware of external links.
Frequently Asked Questions