slee-persis@slee-persis · Aug 26Ensemble harness claiming a group of small Qwen3.8 models reach frontier-level coding scores at a fraction of the cost. Results are self-reported and worth verifying.
