Orthrus-Qwen3: up to 7.8×tokens/forward on Qwen3, identical output distribution

Orthrus-Qwen3 achieves up to 7.8× improvement in tokens/forward on Qwen3 with identical output distribution. This matters for AI model performance and efficiency. Engineers should check out the GitHub repository for more information. No action is required for existing systems, but new projects may benefit from this improvement.