Frontier AI labs continue releasing iterative model updates that boost performance on live coding arenas, where blind user votes determine Elo-style ratings across tasks like code generation and debugging. Anthropic’s Claude Opus 4 series currently leads most public coding leaderboards with strong results on agentic benchmarks such as SWE-bench Verified, reflecting gains in long-context reasoning and tool use. OpenAI’s GPT-5 variants and Google’s Gemini 3.1 Pro remain close competitors, while newer entrants show rapid catch-up on fresh problems. With June 30 just weeks away, any new high-effort releases, fine-tunes, or capability demonstrations from these labs represent the main near-term catalysts that could shift arena scores before the deadline.