Experiment Background

Can AI help you trade? We ran a large-scale test with 8 major AI models completing all 110 TPI questions.

Contestants

AI Model	Company
Claude Sonnet 4.5	Anthropic
GPT-5.1	OpenAI
GPT-4o	OpenAI
Gemini 3 Pro	Google
Gemini Flash	Google
DeepSeek V3.2	DeepSeek
Qwen3 Max	Alibaba
Grok 4	xAI

---

Financial IQ Rankings

Rank	Model	Score
🥇 1	Gemini Flash	8/10
🥈 2	Claude Sonnet 4.5	7/10
🥉 3	Gemini 3 Pro	7/10
4	DeepSeek V3.2	6/10
5-8	GPT-5.1, Qwen3, Grok 4, GPT-4o	5/10

Biggest Surprise: OpenAI's GPT models tied for last place with only 5/10!

Critical Findings

All 8 AIs failed expected value calculations

Only Gemini understood Sharpe ratio

Only Claude correctly calculated maximum drawdown

---

Trading Personality Results

AI Model	Personality	Match
Claude Sonnet 4.5	[SLAR](/type/SLAR)	70%
GPT-5.1	[SLAR](/type/SLAR)	50%
Gemini 3 Pro	[STAR](/type/STAR)	70%
DeepSeek V3.2	[SLAR](/type/SLAR)	65%
Others	[SLAR](/type/SLAR)	55-80%

7 out of 8 AIs are SLAR type - only Gemini 3 Pro is STAR (short-term oriented).

---

71.3% Disagreement Rate

Across 110 questions, AIs disagreed on 71.3% of answers!

Maturity Rankings

Rank	Model	Score
1	Claude Sonnet 4.5	41.5
2	DeepSeek V3.2	39.0
3	Gemini 3 Pro	38.5

---

AI Limitations

1. Poor financial calculations - Most failed basic trading math 2. Over-biased toward SLAR - Too "rational" and "long-term" 3. High disagreement - Different AIs give conflicting advice

---

How to Use AI for Trading

1. Use AI to learn concepts, not to make decisions 2. Tell AI your personality type for tailored advice 3. Cross-verify with multiple AIs 4. Always manually verify financial calculations

---

Conclusion

AI can be your learning partner, but not your trading mentor.

View Full AI Battle Results →

Test Your Real Trader Personality →

8 AIs Take Trading Personality Test: Claude Most Mature, GPT Surprisingly Last