Anthropic launches Claude 4.5 Opus with multi-agent orchestration

Anthropic has released Claude 4.5 Opus, its most capable model yet, featuring native multi-agent orchestration, dynamic tool synthesis and a 500,000-token context window. It scores 74.8% on SWE-bench Verified, beating OpenAI's GPT-6.1 Sol and Google's Gemini 2.5 Ultra, at $15 per million input tokens and $75 per million output tokens.

• 500,000-token context with up to 100% needle-in-a-haystack recall
• 74.8% on SWE-bench Verified, ahead of GPT-6.1 Sol and Gemini 2.5 Ultra
• Dynamic tool synthesis: writes and runs Python or Rust helpers on the fly
• Pricing: $15 per million input tokens, $75 per million output tokens