Claude Haiku 5.5 Matches GPT-6 Luna Rates for Prompts Up to 100,000 Tokens

Anthropic’s Claude Haiku 5.5 and OpenAI’s GPT-6 Luna list identical input and output rates for prompts up to 100,000 tokens, offering a direct price comparison for shorter model requests.

As of Nov. 2, each model’s standard listed rates were $0.10 per million input tokens and $0.50 per million output tokens in that tier. Anthropic launched Haiku 5.5 on Oct. 7.

The match does not cover every prompt length. OpenAI lists those GPT-6 Luna rates for prompts up to 272,000 input tokens. Anthropic lists higher rates for prompts above 100,000 tokens: $0.50 per million input tokens and $2.50 per million output tokens.

A request’s total bill also depends on how many input and output tokens it uses. The listed rates concern inference, or querying a trained model, while training costs cover resources used to build the model.

Across 45 frontier models, the estimated amortized cost of hardware and energy for final training runs grew roughly 2.4-fold annually from 2016. A separate benchmark-based inference-cost estimate found that reaching a given performance level across five benchmarks became about 47% cheaper per quarter from 2023. That measure follows the lowest-priced models meeting benchmark thresholds; it does not mean every API price or real-world task became cheaper at the same rate.

The two pricing tiers offer a comparison for specific models and prompt lengths. They do not show whether lower inference prices offset the costs of training and infrastructure.