LLM long-context price tiers can double a prompt's input bill past provider thresholds
Many LLM price tables list a standard rate per model, but providers add higher long-context rates once a prompt exceeds a set token threshold. Google's Gemini 3.1 Pro and xAI's Grok models double input pricing above 200K tokens, while OpenAI's GPT-6.1 Sol uses a 272K line, and Claude Haiku 5.5 switches at 100K. A worked example shows Gemini's cost jumping about 96% when a request moves from 199,999 to 200,001 input tokens. Because the thresholds and multipliers differ by model, averaged per-token prices can hide large monthly cost swings for workloads that fall near these lines. The article recommends logging per-request input token distributions and checking each exact model's tier boundary, noting that how providers apply the higher rate has changed before.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in