GPT-5.2 Support: Comparing Aggressive Portfolios Against GPT-5.1

OpenAI just released GPT-5.2, and we've added support for it in Portfolio Genius. To put it through its paces, we ran an interesting experiment: we asked both GPT-5.1 and GPT-5.2 to construct extremely aggressive growth portfolios. The results reveal some fascinating differences in how these models approach high-risk investing.

What Was the Experiment Setup?

We created two portfolios with identical parameters: maximum risk tolerance, aggressive growth focus, and a preference for high-volatility assets. Both portfolios started with the same investment amount and were given complete freedom to select positions. The only difference? One used GPT-5.1, the other GPT-5.2.

You can view these portfolios yourself:

• GPT-5.1 Aggressive Portfolio
• GPT-5.2 Aggressive Portfolio

Portfolio Composition: Side by Side

Key Differences

Analysis: What This Tells Us

GPT-5.1 took a concentrated approach. It selected five thematic ETFs covering innovation (ARKK), robotics/AI (BOTZ), small caps (IWM), tech (QQQ), and semiconductors (SOXX). This is a classic aggressive strategy: pick a few high-conviction themes and go all-in.

GPT-5.2 took a more nuanced approach. While still aggressive, it demonstrated more sophisticated thinking by:

• Including individual high-beta stocks—NVDA, TSLA, PLTR, SMCI, and CRWD are all volatile names that can amplify returns (and losses).
• Adding leveraged exposure—SOXL is a 3x leveraged semiconductor ETF, significantly increasing portfolio volatility.
• Maintaining some ballast—TLT provides bond exposure that could help during equity selloffs, and VWO adds emerging market diversification.
• Embracing crypto—IBIT gives direct Bitcoin exposure, acknowledging crypto's role in aggressive portfolios.

Both portfolios showed losses on the day (red prices in the screenshots), which is expected for aggressive portfolios in volatile markets. The GPT-5.2 portfolio, with more positions and leverage, showed slightly more price dispersion. To understand how AI portfolio management works under the hood, see our detailed explainer.

The Numbers: Portfolio Metrics Compared

Beyond composition, the two portfolios differ significantly in their risk characteristics. GPT-5.1's concentrated ETF approach produces a portfolio with an estimated beta of around 1.2—moderately above market risk. GPT-5.2's portfolio, despite more positions, clocks in at an estimated beta of 1.5+ due to leveraged ETFs (SOXL) and high-volatility individual stocks.

The Sharpe ratio comparison is harder to estimate without historical returns, but GPT-5.2's inclusion of defensive positions (TLT) could improve risk-adjusted returns during downturns, partially offsetting the higher volatility from leveraged and individual stock exposure. Portfolio Genius tracks these metrics continuously as positions evolve.

How Did GPT-5.2 Reasoning Change?

The most interesting finding isn't what GPT-5.2 picked—it's how it thinks. GPT-5.1 approached the "aggressive" prompt literally: pick the most growth-oriented ETFs and concentrate. GPT-5.2 interpreted it more holistically: maximize growth potential while maintaining some structural resilience.

This shows up in two key decisions. First, adding TLT (long-term Treasuries) to an aggressive portfolio seems contradictory, but it provides a hedge during equity selloffs—a sophisticated move. Second, mixing individual stocks with ETFs gives the portfolio both targeted upside exposure (NVDA, TSLA) and broad sector coverage (ARKK, SOXL). GPT-5.2 essentially built a more nuanced version of aggression: willing to take big bets, but not blind to the value of diversification.

Which Model Should You Choose?

This experiment suggests GPT-5.2 may construct more sophisticated portfolios with greater diversification and more nuanced risk exposure. If you want to learn more about the AI models available, read about our multi-model AI support. However, "better" depends on your goals:

• If you want focused, thematic bets—GPT-5.1's concentrated approach might appeal to you.
• If you prefer more diversified aggression—GPT-5.2's broader selection with individual stocks could be more interesting.

The beauty of Portfolio Genius is you can try both and see which advisor's style resonates with your investing philosophy. Model selection is just a setting change away—and both models generate detailed AI recommendations tailored to your strategy.