OpenAI Cuts Prices, Lower Token Costs Benefit Applications

English summary

Event: On July 30, OpenAI announced price cuts for two tiers of GPT-5.6 models, citing improved inference service efficiency. Key information: Luna (fastest and cheapest tier): cut about 80%, input/output from $1/$6 to $0.2/$1.2 (per million tokens). Terra (mid-tier): cut about 20%, input/output from $2.5/$15 to $2/$12. Sol (flagship tier): no price cut, remains $5/$30. OpenAI explicitly stated: cheaper Chinese open-source models (Kimi K3, DeepSeek V4, etc.) are increasing pressure on OpenAI/Anthropic, forcing them to justify the high prices; meanwhile, long-horizon agent tasks consume far more tokens than chat, and both companies have been repeatedly adjusting prices and limits in recent months. ...