Reaction
The case for preparing for AI inference without subsidies
A user argues that serving AI models remains costly despite OpenAI and Anthropic’s work to cut inference costs. They expect limit resets and token subsidies to eventually end.
TLDR
A user credits OpenAI and Anthropic with optimizing their models and infrastructure to cut LLM inference costs, but says serving AI models still costs a lot. They urge planning for an end to limit resets and token subsidies, claiming open-weight LLMs can already do 80–90% of the tasks people use closed models for at a fraction of the cost.
Combined views
17
1 Source, first seen 1h ago