DeepSeek V4Pro Leaderboard Launches, Smart Money Betting on DeepSeek Model to Top the Charts This Year, Boosting Odds by 900%

According to PolyBeats monitoring, in the prediction market Polymarket, 20 minutes ago, a savvy investor placed a $4.9k bet on "Did DeepSeek have the #1 AI model by December 31st?" The average buy probability was 73.3%, causing the "Yes" probability to increase from 8.0% to 73.4%.

dry-september also invested $4.9k, with the top relevant category being artificial intelligence, resulting in a net profit of $4.0k. Out of 23 settled trades in this category, the win rate was 17/23 (74%), with 2 trades where the buy price was less than $0.8 and the sell price was above $0.95. Within a similar cost range ($0.651-$0.7), the median historical investment amount was $4.7k.

As per market rules, DeepSeek needs to have the highest model score on the LMArena leaderboard this year.

Based on DeepSeek's own technical report, the V4-Pro-Max data shows that it scored 93.5 on LiveCodeBench, Codeforces rating of 3206, SWE Verified 80.6, GPQA Diamond 90.1, and emphasizes that Pro-Max is the maximum reasoning effort mode. Artificial Analysis also ranks DeepSeek V4 Pro 0813 (Reasoning, Max Effort) as Intelligence Index 2/104, with a score of 53, indicating it is close to the top tier in third-party comprehensive intelligence evaluation.

However, the current LMArena leaderboard still shows that DeepSeek is far from the rule-defined "#1." As of August 12th, the 1st place belongs to Anthropic's claude-fable-5 with a score of 1506±5; the 2nd and 3rd places are also held by Anthropic's Claude Opus series. DeepSeek's newly emerged deepseek-v4-pro-max-20260813 is currently in the AutoEval stage with a score of 1465±10, not yet officially ranked; the formally ranked deepseek-v4-pro is in 50th place with a score of 1458±4.
---------------------------------
See the future sooner, follow @PolyBeats_Bot
See tomorrow, today. Follow @PolyBeatsEN