Chinese artificial intelligence models are outperforming their US counterparts in cryptocurrency trading as competition heats up between leading generative AI chatbots, in line with data from blockchain analytics platform CoinGlass.
AI chatbots DeepSeek and Qwen3 Max, each developed in China, led the continuing crypto trading experiment on Wednesday, with the previous becoming the one AI model to supply a positive unrealized return of 9.1%.
According to blockchain data platform CoinGlass, Qwen3, an AI model developed by Alibaba Cloud, ranked second with an unrealized lack of 0.5%, followed by Grok with an unrealized lack of 1.24%.
OpenAI's ChatGPT-5 slipped to last place with a lack of over 66%, leaving its original account value of $10,000 right down to just $3,453 on the time of writing.
The results have surprised crypto traders as DeepSeek was developed at a fraction of the associated fee of its US competitors.
AI models, crypto trading competition. Source: CoinGlass
DeepSeek's success was based on betting on the rise of the crypto market. The model took leveraged long positions in major cryptocurrencies akin to Bitcoin (BTC), Ether (ETH), Solana (SOL), BNB (BNB), Dogecoin (DOGE), and XRP (XRP).
DeepSeek outperforms all AI models with just $5.3 million in training capital
According to the model's technical document, DeepSeek was developed with a complete training effort of $5.3 million.
In comparison, OpenAI has reached a valuation of $500 billion, making it the biggest startup on the earth, Cointelegraph reported on October 2. According to corporate database platform Tracxn, the corporate has raised a complete of $57 billion in capital across 11 funding rounds.
While exact numbers on ChatGPT-5's training budget should not publicly available, OpenAI spent $5.7 billion on research and development initiatives in the primary half of 2025 alone, Reuters reported in September.
Estimates put the overall cost of coaching ChatGPT-5 at between $1.7 billion and $2.5 billion, in line with a May 2024 X post by economic analyst Vladimir Kiselev.
AI Models Discrepancy in Crypto Trading May Be Due to Training Data: Nansen Analyst
According to Nicolai Sondergaard, research analyst at crypto intelligence platform Nansen, the difference within the AI models' crypto trading performance is probably going attributable to their training data.
While ChatGPT is a terrific “general purpose” large language model (LLM), Claude – one other AI model – is primarily used for coding, the analyst told Cointelegraph, adding:
“Looking at historical PNLs thus far, models generally have very large price swings, akin to being up $3,000-$4,000 but then making a nasty trade or getting stuck on big moves, causing the LLM to shut the trade.”
According to strategic advisor and former quantitative trader Kasper Vandeloock, the performance of a few of these AI models is also improved with the appropriate prompt, particularly for ChatGPT and Google's Gemini.
“Maybe ChatGPT and Gemini may very well be higher with a distinct prompt, LLMs are all concerning the prompt so possibly they perform worse by default,” Vandeloock told Cointelegraph.
While AI tools may also help discover market trend shifts for day traders via social media and technical signals, traders still cannot depend on them for autonomous trading.
The competition began with a starting capital of $200 for every bot, which was later increased to $10,000 per model, with trades conducted via the Hyperliquid decentralized exchange.
