While the open approach is commendable. The 32b and 36b models are inferior to qwen 3.8 27b, at least according to benchmark numbers. I would have liked to have

While the open approach is commendable. The 32b and 36b models are inferior to qwen 3.8 27b, at least according to benchmark numbers. I would have liked to have seen both compared to 27b, not only the dense one. Also would have liked to see more coding benchmarks in the full table.