The Biggest Model Lost! Claude Coding Challenge EP04

Which Claude model writes the best code? Episode 4 of the LLM Coding Challenge — this time we build a word frequency counter with a horizontal bar chart. Same prompt. One shot. No follow-ups. Each model writes a Python program that counts word frequencies in sample text and displays the top 5 most frequent words as a horizontal bar chart using # characters. We score each model across 8 criteria: - Correct word counting - Top 5 display - Bar chart with # characters - Scaled bar lengths - Sample text included - Docstring included - Clean code structure - Main guard included Results: - Opus: 7/8 — Most concise (30 lines), clean output with scaled bars, but no docstring - Sonnet: 8/8 — Most thorough (96 lines), full documentation, two functions, prints raw frequency data - Haiku: 7/8 — Middle ground (49 lines), good docs and structure, but bars use raw counts instead of scaling Sonnet wins for the first time! All code was generated using Claude Code CLI with the -p (one-shot) flag. Source code: https://github.com/GoCelesteAI/llm_coding_challenge --- Chapters: 0:00 Intro 0:15 The Prompt 0:30 Opus 2:06 Sonnet 4:04 Haiku 5:52 Scorecard 6:27 Outro --- #Claude #AI #CodingChallenge #LLM #Opus #Sonnet #Haiku #Python #Programming #AICode #WordFrequency #BarChart Tags Claude, AI, coding challenge, LLM, Opus, Sonnet, Haiku, Python, word frequency, bar chart, programming, AI code, Claude Code, Anthropic, code comparison, AI models, Claude Code CLI