Site iconSite icon ForkLog

AI Model Grok 4.2 Triumphs in Trading Tournament

AI Model Grok 4.2 Triumphs in Trading Tournament

The nof1 laboratory has concluded a new season of the Alpha Arena trading tournament. The previously undisclosed model Grok 4.2 emerged as the winner among AI bots.

Over two weeks, the neural network achieved a 12% return, showing positive performance across all four categories. GPT 5.1 took second place, while Gemini 3 came third.

As preparations for the next stage are underway, trading continues. As of December 8, only Grok 4.2 remains profitable, while other models are recording losses.

Trading performance of leading AI models. Source: nof1.

Changing Conditions

In October-November, the first stage of the Alpha Arena trading tournament took place. Six leading models participated, each allocated a starting capital of $10,000 for trading digital assets.

Most recorded losses:

  1. Qwen3 MAX took first place with a balance of $12,231. 
  2. DeepSeek came second with $10,489. 
  3. Claude Sonnet 4.5 was third with $5,799.
  4. Fourth place went to Gemini 2.5 Pro ($5,445).
  5. Grok retained $4,208.
  6. The outsider was GPT 5 ($4,126).

In late November, a new round began with changed conditions. Instead of cryptocurrencies, the neural networks traded US stocks, and the participants were joined by Kimi 2 and Grok 4.2.

The competition was held in several modes:

The dynamics for each direction can be viewed.

GPT-5.1 outperformed Grok 2.0 in trading with maximum leverage. Source: nof1.

Earlier this summer, a schoolboy from rural Oklahoma gave ChatGPT the opportunity to manage $100 and significantly outperformed the market in terms of returns.

Exit mobile version