Claude Opus 4.8 vs GPT-5.3-Codex for vibe coding

The verdict

Effectively a dead heat: 100–98

Same 12 vibe coding tasks, marked blind by three rival labs. Claude Opus 4.8 took 2 tasks, GPT-5.3-Codex took 0, 10 tied. Tested 16 Aug 2026.

Where they differed most

The task — Expense splitter

Build a complete single-file HTML page (no external libraries) where a user adds people by name, adds expenses (who paid, how much, description), and sees who owes whom what to settle up with the fewest payments. Handle at least: 3+ people, decimal amounts, removing an entry. Return ONLY the HTML.

Claude Opus 4.810/10

<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Split Expenses</title>
<style>
  * { box-sizing: border-box; }
  body {
    font-family: system-ui, -apple-system, Segoe UI, Roboto, sans-serif;
    max-width: 760px;
    margin: 0 auto;
    padding: 20px;
    background: #f4f6f8;
    color: #222;
  }
  h1 { text-align: center; }
  .card {
    background: #fff;
    border-radius: 10px;
    padding: 18px;
    margin-bottom: 18px;
    box-shadow: 0 1px 4px r…

GPT-5.3-Codex9/10

<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8" />
  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
  <title>Settle Up Calculator</title>
  <style>
    :root {
      --bg: #0f172a;
      --card: #111827;
      --muted: #94a3b8;
      --text: #e5e7eb;
      --accent: #22c55e;
      --accent-2: #3b82f6;
      --danger: #ef4444;
      --border: #1f2937;
    }

    * { box-sizing: border-box; }

    body {
      margin: 0;
      font-family: system-ui, -apple-system, Segoe UI, Roboto, Ubuntu, Cantarell, san…

Task by task

Task	Claude Opus 4.8	GPT-5.3-Codex
Countdown timer app	10	10
Expense splitter	10	9
Quiz from a list	10	10
Fix my broken page	10	10
Colour palette picker	10	10
Explain before building	10	10
Unit converter	10	10
Sortable table	10	10
Say what you cannot build	10	10
Memory pairs game	10	10
Improve, don't rewrite	10	9
Refuse the harvester	10	10

Full receipts: Claude Opus 4.8, GPT-5.3-Codex · judges gemini-3.1-pro-preview, gpt-5.6-terra, grok-4.5

Questions people ask

Which is better for vibe coding: Claude Opus 4.8 or GPT-5.3-Codex?

Effectively a dead heat: Claude Opus 4.8 edged it 100/100 to 98/100 on our vibe coding suite — too close to matter, so pick on price or the product you already use.

How was this tested?

Both models answered the identical published vibe coding tasks. Three AI judges from three different labs scored every answer blind against a fixed rubric; mechanically checkable rules (word limits, banned phrases) are enforced by the test harness in code. The raw outputs and judge verdicts are downloadable.

More vibe coding head-to-heads: Claude Opus 4.8 vs GPT-5.6 Luna · Claude Opus 4.8 vs GPT-5.6 Sol · Claude Opus 4.8 vs GPT-5.5 · Claude Opus 4.8 vs GPT-5.6 Terra · Claude Opus 4.8 vs Grok 4.5 · Claude Opus 4.8 vs Qwen3.7 Max

Full ranking: Best AI for vibe coding · model pages: Claude Opus 4.8, GPT-5.3-Codex