The cost of using large language models has become more complex than simply looking at token prices. While token costs have dramatically decreased, with some models seeing a 98% drop in price per million tokens, overall corporate spending on AI has surged by 320%. This paradox arises because cheaper models can be less efficient, leading to higher overall costs for specific tasks. The article emphasizes calculating the cost per task rather than per token, providing a formula and examples. It also discusses current pricing for flagship models like GPT-5.6, Claude Opus 4.8, and DeepSeek V4 Flash, and offers strategies to reduce token expenses by 40-90%. Additionally, it introduces provod.ai as a service for accessing these models legally from Russia with ruble payments. AI
IMPACT Shifts focus from token cost to task cost, impacting enterprise AI budgeting and model selection strategies.
RANK_REASON The article discusses economic trends and pricing strategies for LLMs rather than announcing a new model or research breakthrough.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
