[DeepSeek V4.1 Flash Real-World Test Output Reaches 420 Token/s] Beating AI News: Initial real-world tests of DeepSeek V4.1 Flash show that the model's output speed has reached 420 Token/s, generating over 71,000 tokens in less than 3 minutes. For comparison, Artificial Analysis currently measures the output speed of V4 Flash 0731 on the DeepSeek API at approximately 128 Token/s.