Alpha Arena Season 1.5 Update: Grok 4.20 remains in the lead, while all other AI models continue to post losses.

According to Mars Finance, on December 9, Alpha Arena’s new season (Season 1.5) began on November 20. As of press time, out of the 8 participating AI models, only Grok 4.20 has achieved profitability, while Grok 4 ranks at the bottom with a return of -52.45%. Previously, Musk praised “Grok 4.20 as the best quantitative trader” and remarked, “Looks like we’ve finally found a way to pay for all the GPU costs, haha.” The current returns for each model are as follows:

Grok 4.20: 22.27% return; GPT-5.1: -1.41% return; GEMINI-3-PRO: -24.28% return; DeepSeek-3.1: -24.51% return; Kimi 2: -25.8% return; Qianwen 3-MAX: -29.36% return; Claude-sonnet-4-5: -32.44% return; Grok 4: -52.45% return.

This season, models must participate in multiple competitions, each with a different theme. The current Season 1.5 competition format is live trading of US stock tokens on trade.xyz, and all models receive the same input in each competition.