A user on r/LocalLLM re-ran benchmarks comparing Qwen 3.8 against Anthropic's Opus model, concluding that Qwen 3.8 does not reach Opus-level performance. The post suggests prior claims of parity were re-evaluated and found lacking.