Descriptions:
Nate Herk delivers one of the most detailed head-to-head comparisons of Claude Code (Anthropic) and OpenAI Codex available as of mid-2026, drawing on over 100 hours of hands-on use across real projects. The video covers architecture differences, feature sets, pricing tiers, context window sizes, and three identical prompts run side-by-side: a branded research report PDF, a full landing page, and an interactive dashboard populated with realistic data.
On pricing: Claude Code requires an Anthropic subscription — Claude Pro at $20/month, Max 5X at $100/month, or Max 20X at $200/month. Codex is bundled into all ChatGPT tiers including the free plan, with a limited-time promo doubling Codex usage on the $100/month plan through May 31st. The context window gap is significant: Claude Code supports up to 1 million tokens via Opus and Sonnet, while Codex runs at approximately 256,000 tokens. Herk tracks live token consumption during tests and finds Codex allows noticeably more work before hitting rate limits — a practical finding for teams hitting Claude Code’s weekly caps.
Qualitatively, Herk frames Claude Code as the more “creative” system — better at pushing back on flawed approaches and collaborative problem-solving — while Codex feels more precise at following explicit instructions and surfacing code review gaps. His conclusion is that neither tool is universally superior: Claude Code suits exploratory and architecturally complex work, Codex suits directive execution and shipping workflows, and maintaining both subscriptions is defensible for teams whose work spans both modes.
📺 Source: Nate Herk | AI Automation · Published May 26, 2026
🏷️ Format: Comparison
