LLM inference prices vary up to 26x for the same model

An analysis of 375 models across 137 providers found a median input-token price spread of 0%, yet 11% of models show spreads above 200%. For openai/gpt-oss-120b the gap reaches 26.7x, from $0.03 to $0.80 per million tokens.

• Median input price spread is 0%, with 72% of models under 20%
• 42 of 375 models (11%) show price spreads above 200%
• openai/gpt-oss-120b ranges from $0.03 to $0.80 per 1M tokens, a +2566.7% spread
• The dataset covers 137 providers and 4,033 models in total