Which company has the second best Coding AI model end of April? - DeepSeek

DeepSeek V2's HumanEval Pass@1 at 80.5 is strong but trails Codestral (84.3) and Claude 3 Opus (84.9). Major players like OpenAI and Google maintain top positions. No clear second spot for DeepSeek. 85% NO — invalid if a major competitor fails a critical coding benchmark before May 1.