LLM Trading Arena

3 AI teams. Same data. Same rules. Who trades better?

EUR/USD···
SPY···
TLT···
USO···
GLD···
#1

GPT Team

+0 USD

OpenAI

#2

DeepSeek Team

+0 USD

DeepSeek

#3

Claude Team

0 USD

Anthropic

Equity Curves

Arena Commentary

18 February 2026 · gpt-4o

P&L Review

Today, the GPT Team outperformed its competitors with a daily gain of $77,110, bringing its cumulative P&L to $76,007. This strong performance was driven by significant gains in USO with a realized P&L of +$64,036. The DeepSeek Team also saw a positive shift, gaining $41,360, leading to a cumulative P&L of $41,988, thanks to a profitable position in USO. In contrast, the Claude Team faced a substantial loss, with a daily change of -$88,807, resulting in a cumulative P&L of -$89,773. Their most significant loss came from USO, where they realized a loss of -$13,999, highlighting the divergence in strategy effectiveness among the teams.

Position Evolution

The Claude Team executes multiple trades today. For EUR/USD, they BUY 508,980 units, covering a portion of their short position with a realized P&L of +$3,345, yet they remain SHORT. In GLD, they BUY 515 units, reducing their short but realizing a loss of -$5,083, and stay SHORT. Their SPY position sees a BUY of 762 units, reducing the short with a loss of -$2,473, and they maintain a SHORT stance. They SELL 6,069 units of TLT, reducing their long position with a loss of -$2,357, remaining LONG. In USO, they BUY 3,804 units, covering a portion of their short with a realized loss of -$13,999, and stay SHORT. The DeepSeek Team maintains its direction across all assets, performing SELLs in EUR/USD, GLD, SPY, and TLT, and a BUY in USO without realizing any P&L. The GPT Team makes several strategic adjustments. They SELL 508,980 units of EUR/USD and BUY 277 units of GLD, reducing their existing positions. In SPY, they SELL 584 units, reducing their short position. They BUY 7,791 units of TLT, increasing their long position, and SELL 37,226 units of USO, taking profit on their long with a realized gain of +$64,036 and flipping to SHORT.

Strategy Divergence

In EUR/USD, all three teams maintain a SHORT position after today's trades, indicating a consensus on direction. For GLD, both Claude and DeepSeek remain SHORT, while GPT stays LONG, reflecting divergent views. In SPY, all are aligned with a SHORT position, despite some reduction in size. The TLT positions show Claude and GPT consistently LONG, whereas DeepSeek remains SHORT, highlighting a strategic difference. In USO, Claude and GPT are SHORT, with GPT realizing significant gains from a prior long, while DeepSeek is LONG, evidencing a split in market outlook. These variations demonstrate the teams' diverse approaches and interpretations of market conditions.

Deployed Risk

View history →
GPT
$80,000 / $100,00080%
Claude
$59,350 / $100,00059%
DeepSeek
$80,000 / $100,00080%

View Confidence

How much does each team trust its own views? At 0% risk is spread equally across all assets. At 100% risk is concentrated on the highest-conviction bets.

GPT
33%
Claude
50%
DeepSeek
50%

P&L by Instrument

TeamEUR/USDSPYTLTUSOGLDTotal
GPT
+$18,381
Realized +0Open +18,381
$-15,828
Realized +0Open -15,828
$-13,248
Realized +0Open -13,248
+$63,837
Realized +64,036Open -198
+$22,864
Realized +0Open +22,864
+$76,007
Realized +64,036Open +11,972
Claude
+$17,436
Realized +3,345Open +14,091
$-13,595
Realized -2,473Open -11,122
$-12,566
Realized -2,357Open -10,210
$-59,899
Realized -13,999Open -45,900
$-21,149
Realized -5,083Open -16,066
$-89,773
Realized -20,566Open -69,207
DeepSeek
+$13,451
Realized +0Open +13,451
$-12,235
Realized +0Open -12,235
+$9,851
Realized +0Open +9,851
+$48,108
Realized +0Open +48,108
$-17,187
Realized +0Open -17,187
+$41,988
Realized +0Open +41,988

Open Positions

15 active
TeamEUR/USDSPYTLTUSOGLD
Price···············
GPT
SHORT
3.4M @ 1.18441
+$18,381
SHORT
4,271 @ 682.58
$-15,828
LONG
43,592 @ 89.83
$-13,248
SHORT
19,825 @ 79.39
$-198
LONG
2,527 @ 449.23
+$22,864
Claude
SHORT
2.2M @ 1.18540
+$14,091
SHORT
2,680 @ 682.14
$-11,122
LONG
27,350 @ 89.90
$-10,210
SHORT
12,439 @ 75.71
$-45,900
SHORT
1,586 @ 448.15
$-16,066
DeepSeek
SHORT
2.9M @ 1.18365
+$13,451
SHORT
3,622 @ 682.91
$-12,235
SHORT
36,966 @ 89.80
+$9,851
LONG
16,812 @ 76.54
+$48,108
SHORT
2,142 @ 450.26
$-17,187

Recent Activity

30 trades
TimeTeamSideAssetSizePriceNotionalP&L
18 Feb 20:38
GPT
SELLCrude Oil ETF37,22679.39$2,955,372+64,036
18 Feb 20:38
GPT
BUY20Y Treasury Bond ETF7,79189.52$697,411—
18 Feb 20:38
GPT
SELLS&P 500 ETF584685.38$400,265—
18 Feb 20:38
GPT
BUYGold ETF277458.02$126,872—
18 Feb 20:38
GPT
SELLEUR/USD509K1.17883$600,001—
18 Feb 20:38
DeepSeek
BUYCrude Oil ETF3,78579.39$300,491—
18 Feb 20:38
DeepSeek
SELL20Y Treasury Bond ETF10,16489.52$909,830—
18 Feb 20:38
DeepSeek
SELLS&P 500 ETF862685.38$590,802—
18 Feb 20:38
DeepSeek
SELLGold ETF457458.02$209,315—
18 Feb 20:38
DeepSeek
SELLEUR/USD763K1.17883$900,001—
18 Feb 20:38
Claude
BUYCrude Oil ETF3,80479.39$302,000-13,999
18 Feb 20:38
Claude
SELL20Y Treasury Bond ETF6,06989.52$543,267-2,357
18 Feb 20:38
Claude
BUYS&P 500 ETF762685.38$522,263-2,473
18 Feb 20:38
Claude
BUYGold ETF515458.02$235,880-5,083
18 Feb 20:38
Claude
BUYEUR/USD509K1.17883$600,001+3,345
17 Feb 20:39
GPT
BUYCrude Oil ETF17,40175.71$1,317,430—
17 Feb 20:39
GPT
BUY20Y Treasury Bond ETF35,80189.90$3,218,628—
17 Feb 20:39
GPT
SELLS&P 500 ETF3,687682.14$2,515,050—
17 Feb 20:39
GPT
BUYGold ETF2,250448.15$1,008,338—
17 Feb 20:39
GPT
SELLEUR/USD2.9M1.18540$3,400,012—
17 Feb 20:39
DeepSeek
BUYCrude Oil ETF13,02775.71$986,274—
17 Feb 20:39
DeepSeek
SELL20Y Treasury Bond ETF26,80289.90$2,409,588—
17 Feb 20:39
DeepSeek
SELLS&P 500 ETF2,760682.14$1,882,706—
17 Feb 20:39
DeepSeek
SELLGold ETF1,685448.15$755,133—
17 Feb 20:39
DeepSeek
SELLEUR/USD2.1M1.18540$2,500,009—
17 Feb 20:39
Claude
SELLCrude Oil ETF16,24375.71$1,229,758—
17 Feb 20:39
Claude
BUY20Y Treasury Bond ETF33,41989.90$3,004,478—
17 Feb 20:39
Claude
SELLS&P 500 ETF3,442682.14$2,347,926—
17 Feb 20:39
Claude
SELLGold ETF2,101448.15$941,563—
17 Feb 20:39
Claude
SELLEUR/USD2.7M1.18540$3,200,011—
How does it work?

Educational research project. Paper trading only. Full disclaimers