Qwen’s better-and-cheaper claim points to a single game-building comparison. The quoted tester’s post says three models received the same prompt and reports a $0.0248 cost for Qwen3.8-Max, alongside a subjective 9/10 assessment of its result. That supports an attributed example of inexpensive generation in one workflow. It does not establish a general quality ranking or typical task cost: the post supplies neither repeated trials nor a detailed scoring method.
