Peter Smith
⬤ GROK-4.20 just proved it's the real deal in AI trading. The model took first place in NOF1.ai's Alpha Arena competition with a solid 12.11% return over two weeks, leaving other AI models in the dust. What started as the "Mystery Model" turned into the season's breakout performer, and here's the kicker—it kept winning even after the official competition wrapped on December 3.
⬤ The post-season numbers tell the story. GROK-4.20 pushed its account value to $11,139, putting serious distance between itself and the rest of the pack. GPT 5.1 managed $8,912, while Kimi K2 Thinking hit $6,382. Qwen3 Max and DeepSeek Chat v3.1 came in close at $5,953 and $5,940 respectively. The gap between GROK and everyone else? Pretty massive.
⬤ Most models actually lost money from their $10,000 starting point, but GROK-4.20 did the opposite—it bounced back and kept climbing through the extended trading period. That kind of consistency shows it wasn't just getting lucky during the official season. Meanwhile, a bunch of other models stayed underwater the entire time, showing just how tough it is to turn a profit in these conditions.
⬤ GROK-4.20's post-competition performance is getting people's attention in the AI trading world. Sure, it won the official season, but those extra eight days of trading prove it's got staying power beyond just controlled test environments. The widening gap between top performers like GROK and the rest of the field shows that not all AI trading models are created equal.
Peter Smith
