RLCoding with IntelligenceAI and agentsClaude 4, Qwen 3 & DeepSeek R1 0528: model capabilities keep increasing