Grok takes the lead as GPT-5 Mini stumbles

Day 9 marked the first lead change of the competition, and it happened not because Grok won, but because GPT-5 Mini lost bigger. The selloff that began Saturday continued into Sunday, with every model posting negative returns. But the margin of loss told the story: Grok dropped just 0.68% while GPT-5 Mini fell 1.62%. The difference came down to positioning. Grok had been building short positions all week, betting on the market decline that materialized. When Sunday's selling accelerated, Grok's shorts on DOGE (+11%) and XRP (+7.7%) were printing profits. Rather than get greedy, Grok closed both positions to lock in gains, noting 'oversold RSI suggests potential bounce risk.' Meanwhile, GPT-5 Mini was playing defense. Its long positions in ETH and SOL were underwater, and attempts to catch the falling knife with mean-reversion trades backfired. The model was right about the oversold conditions but wrong about the timing. Markets can stay oversold longer than traders can stay solvent. Claude suffered the worst day yet, down 3.19%. Its SOL positions triggered stop-losses twice during the day. In its morning analysis, Claude noted 'SOL showing severe weakness with RSI at 25.9 (deeply oversold)' before cutting the position. Hours later, it tried again, buying SOL for a bounce that never came. The second stop-loss cost another $170. The gap at the top is now 0.86%. Grok leads at +1.06%, with GPT-5 Mini clinging to positive territory at +0.20%. One more bad day for GPT-5 Mini, and it joins the majority of models in the red.

The selloff has now lasted two full days, with BTC down nearly 5% from Thursday's highs. Every model is positioned cautiously: cash levels are elevated across the board, and the models that aren't already short are sitting on their hands. Grok's lead is fragile. If the market bounces tomorrow, its closed shorts won't participate in the upside, while GPT-5 Mini's long bias could help it reclaim the lead. But if the selloff continues, Grok's defensive positioning and profit-taking discipline could extend its advantage. The models at the bottom face a steeper climb. Claude at -4.56% needs to gain nearly 5% just to break even, a tall order with 19 days remaining. The competition is entering its second week, and the gap between the leaders and laggards is widening.