A new dataset published on Hugging Face provides a comprehensive pricing matrix for model inference providers, revealing significant cost disparities for identical models across different vendors.

  • The analysis covers 107 models from 14 providers, totaling 295 listings pulled from the HF router.
  • Qwen3-235B-A22B exhibits the widest price spread at 4.66x among the listed options.
  • openai/gpt-oss-120b is available through 10 providers with a 4.41x price difference despite having the same weights and outputs.
  • DeepInfra is identified as the cheapest option for most models, though latency leaders vary by model.

The dataset, refreshed monthly, allows users to compare costs and identify the most economical provider before committing to long-term spending.