ERNIE 4.5 sweeps GPT-4.5 in AI chess showdown
ERNIE 4.5 just swept GPT-4.5 in a three-game chess showdown, winning all three matches decisively. The model demonstrated tighter instruction-following and sharper positional decision-making, even sealing the final match with a 10-move checkmate. It is a reminder that frontier logic tests do not always belong exclusively to OpenAI.
Chess has always been the ultimate sanity check for LLM logic and spatial reasoning.
In a recent face-off shared by X user Md Riyazuddin, Baidu's ERNIE 4.5 took on OpenAI's GPT-4.5 in a three-game match. The outcome was a decisive 3-0 sweep for ERNIE 4.5.
A complete sweep and a 10-move mate
According to the test, ERNIE 4.5 dominated the board right from the opening moves. Match commentary even described the contest as an adult playing a child.
Beyond demonstrating sharper decision-making under pressure and tighter instruction-following, ERNIE 4.5 closed out the final game with a clean 10-move checkmate. The tester also highlighted that ERNIE delivered this performance at a fraction of GPT-4.5's cost.
Why it matters
General-purpose language models are notorious for failing at chess rules, often hallucinating illegal board states after just a few turns. Dominating an opponent over multiple matches reflects robust state tracking and multi-step reasoning.
While a three-game matchup is a small sample size, we think it makes an undeniable statement. ERNIE 4.5 is proving it can run circles around Western frontier models in strict logic tasks, and doing so with a dramatically cheaper price tag.
Sources (1)
- x.com(opens in a new tab)Originalx.com