Can the latest AI DeepSeek beat ChatGPT? Verifying AI capability with trade analysis.

What is DeepSeek?

In January 2025, the latest large language model “DeepSeek R1” released by Chinese AI company DeepSeek has been making waves worldwide, topping app rankings.

■Differences from existing AIs

DeepSeek R1 employs technology that can run deep learning without relying on expensive NVIDIA GPUs, delivering high language processing capabilities even in low-cost environments. Development cost is reported to be only $5.5 million, highlighting its high cost-performance.

■Free to use

DeepSeek R1 is a powerful AI that can be used for free, with strong multilingual accuracy. While many existing AIs offer free options, they are often one generation behind; DeepSeek R1 boasts accuracy at the same level as the latest models. This combination of being free and equipped with cutting-edge technology is a major advantage.

■Cautions about DeepSeek

There may be biases in some data due to political regulations in China, and there can be aspects that show bias. Evaluation criteria can be strict, giving a somewhat stern impression. Functions are somewhat limited and customization is lacking, but the strength lies in simple, clear judgments. Data is stored within China, so usage may require caution.

DeepSeek AI Judgment Results

One-Click FX TrainingWe will load the trade data output by the AI judgment feature into DeepSeek R1 and view the results. Note that in DeepSeek’s validation, chart data cannot be loaded; only trade history is evaluated.

■Result explanation

Total profit/loss, win rate and loss rate, total number of trades, number of profitable trades and win rate, number of losing trades and loss rate, average profit and average loss, maximum profit and maximum loss, analysis by order type, performance by time of day, impact of swaps, correlation between volume and profit/loss, and tendencies in short-term trading are aggregated and, in addition, improvements and an overall assessment are output.

■Judgment points

Although the trade count is shown as 80, in reality there are 103. We are testing various AIs, not just DeepSeek, and with free AIs, the data counts can be inaccurate. There may be some restrictions, but with small data sets, normal results are typical. There are no inconsistencies in improvements or the overall assessment, and the content is not drastically off.

There may be some limits, but with small data sets, normal results are typical. Also, there are no inconsistencies in improvements or overall assessment, and the content does not deviate greatly.

■Overall evaluation

The unfortunate part is that there is no feature to display graphs or images. Explanations are textual only, which reduces persuasiveness.

Also, the judgment “Trade evaluation: 30 points out of 100” is strict. It may clearly state, for example, “risk management is completely lacking, losses snowball,” or “win rate is too low.”

This is test data, but in serious trading, overly harsh feedback might hurt.

---

3. ChatGPT results and comparison with DeepSeek

Using the same trading data, we will also analyze with paid ChatGPT Plus. Here, both chart data and trade history can be loaded, allowing evaluation that combines market fluctuations with trade history.

■ChatGPT result explanation

A chart plotting trade points on the GOLD price movement is displayed, showing entry and exit timings. There is a profit-loss distribution graph and explanation. Key considerations and improvement points are concisely written. Additional analysis information is displayed at the end.

■Differences from DeepSeek

While ChatGPT analyzes the chart data and explains why a large loss at the same point might have occurred due to “bad entry timing, selling during an uptrend,” DeepSeek results state “short-term trade failure: unable to withstand price fluctuation after about 10 hours of holding.” This indicates less thorough analysis. However, given that ChatGPT analyzes chart data while DeepSeek analyzes only trade data, this difference is understandable.

■ChatGPT’s superiority

The obvious difference is the use of visuals with charts, making explanations easier to understand. Considerations are accurate, and there are no issues with content. Feedback is designed to avoid discouraging motivation and provides constructive suggestions for improvement.

Even for such trades, the scoring is around 60 points, and at the end, a note like “If you implement these improvements, you could aim for 80 points or higher!” motivates you to improve.

AI personalities

This trade results from automated trading based on moving average crossovers and trailing stops, with 103 executions, a win rate of 41.7%, total profit/loss of -32,828 yen, PF: 0.87, RR: 1.22, which are not particularly strong results. Therefore, there are many improvement points.

DeepSeek is gaining attention, but saying it is equivalent to paid ChatGPT Plus is an exaggeration. ChatGPT can output images and graphs and can even create videos, which is a clear advantage. However, considering it is free to use, DeepSeek’s performance is excellent.

■AI also has personality.

AI has personality, and various cultures and values may be reflected in it.

In Chinese culture, direct communication is common, so DeepSeek’s strong expressions are understandable. Gemini tends to be logical and somewhat superior in tone, while ChatGPT offers constructive feedback with encouragement. Domestic AIs are also being developed, and it would be interesting if there were a Japanese AI tuned to Japanese preferences with famous strong support.

We have linked reports of AI judgments for DeepSeek and ChatGPT, so please check the links if you want to see detailed content.

ChatGPT:ai_sample_deepseek_vs_chatgpt_01
DeepSeek:ai_sample_deepseek_vs_chatgpt_02

Completely risk-free trading simulator for free practice and verification!

Details page for One-Click FX Training MAX