What happened
Anthropic's Claude Code Max plan does not incur a $5,000 monthly compute loss per user. This figure reflects Anthropic's retail API pricing for Opus 4.6 ($5/$25 per million tokens), not its actual inference costs. Comparable open-weight models, like Qwen 3.5 and Kimi K2.5 on OpenRouter, cost approximately 10% of Anthropic's API rates ($0.39/$2.34 and $0.45/$2.25 per million tokens). This indicates Anthropic's actual compute cost for a heavy user is closer to $500.
Why it matters
This clarifies large language model inference economics, challenging the narrative that AI inference is unprofitable. Procurement teams and founders must differentiate retail API pricing from actual compute expenditure; API rates can inflate perceived operational overhead tenfold. Understanding this mechanism prevents overestimating the financial burden of deploying frontier models, enabling accurate budget forecasting and strategic investment. Open-weight model pricing on platforms like OpenRouter provides a metric for assessing true inference costs, influencing build-versus-buy decisions.
Subscribe for Weekly Updates
Stay ahead with our weekly AI and tech briefings, delivered every Tuesday.
