Ranked 9th in GPT, DeepSeek overtakes Claude, with the strongest trading agent being the outdated 4.6 According to TechFlow, on May 13th, the SMART Benchmark, an AI trading agent benchmark evaluation initiated by Manic.Trade, has officially concluded. This evaluation has completed a total of 387 evaluations, covering 180 testers, 46 models, and 15+model families, making it the most comprehensive trading agent horizontal evaluation in terms of model coverage. It mainly revolves around five dimensions: real-time data processing, multi-source intelligence acquisition, market analysis, trading decision-making, and risk control. The results showed that the Claude family ranked first with 75.9 points, DeepSeek ranked second with 74.4 points, and the GPT family only ranked ninth. In terms of single model, Claude 4.7 Opus has the highest average score of 80.8; DeepSeek V4 Pro ranks second with a score of 78.6; Claude 4.6 Opus ranked third with 78.5 points and contributed the only S-level score of the entire game, reaching a maximum score of 90/100. It is worth noting that DeepSeek scored 15.9 in the Market Analysis dimension, surpassing Claude's 15.4. Manic.Trade stated that the current evaluation shows that the gap in AI trading agents is not mainly reflected in their ability to make trading decisions, but in their ability to integrate multiple sources of information, process verifiable data, and control risks that are closer to real trading execution. At present, users can access Manic.Trade to view complete reports and experience the crypto price prediction market for human traders and AI agents.