Claude Opus 4 LMArena Benchmark Scores

• URL: https://www.threads.com/@purple.soy/post/Da7dusej1-U
• Date Saved: 2026-07-19
• Source: Threads
• Tags: ai-tools, ai-engineering

Summary

@purple.soy posted LMArena (formerly LMSYS Chatbot Arena) benchmark scores for anthropic/claude-opus-4-cumulative, calling it “the best model ever made.”

Scores:

• Overall Arena Score: 84.7
• IF Eval (Instruction Following): 88.2
• Style Control: 83.8
• Creative Writing, editing: 80.7
• Math: 77.7
• Coding: 73.2
• Intelligence (Hard prompts): 67.4
• Long Context: 55.4

Notes:

• Claude Opus 4 is a paid, closed-weight proprietary model from Anthropic
• Available via Anthropic API, AWS Bedrock, Google Vertex AI
• The “cumulative” suffix refers to LMArena’s cumulative vote tracking methodology
• Strong in instruction following and style control; weaker in long context and hard intelligence benchmarks