🚨 OpenAI may have just cut their AI inference costs by more than 50%

Engineers reportedly discovered new internal optimisations that make existing models dramatically cheaper to run:

• More than halved inference costs in early testing • Already applied to logged out ChatGPT traffic • Appears to use existing servers more efficiently • It's not dependent on OpenAI’s new Jalapeño chip • Could eventually unlock higher limits, cheaper APIs or more powerful agents

Would you prefer cheaper API pricing or much higher ChatGPT limits?