GPT-6 Luna vs Claude Haiku 5.5: cost comparison

GPT-6 Luna and Claude Haiku 5.5 have the same list prices per token, but for the same text GPT-6 Luna works out about 1.3× cheaper, because Claude 4.7 and later count roughly 30% more tokens.

GPT-6 Luna OpenAI

$0.10input /1M

$0.01cached /1M

$0.50output /1M

Prompts over 272,000 tokens are billed at $0.20 input / $0.02 cached / $0.75 output per 1M tokens for the whole request.
Batch processing: 50% off input and output ($0.05 / $0.25 per 1M tokens).

Claude Haiku 5.5 Anthropic

$0.10input /1M

$0.01cached /1M

$0.50output /1M

Prompts over 100,000 tokens are billed at $0.50 input / $0.05 cached / $2.50 output per 1M tokens for the whole request.
Uses Anthropic's newer tokenizer, which produces about 30% more tokens for the same text than Claude Sonnet 4.6 and earlier.
Batch processing: 50% off input and output ($0.05 / $0.25 per 1M tokens).

Cost per 1,000 requests

Workload	GPT-6 Luna	Claude Haiku 5.5	Difference
Chatbot reply1,500 input / 400 output tokens	$0.35	$0.455	1.3×
RAG answer8,000 input / 600 output tokens	$1.10	$1.43	1.3×
Agent step30,000 input (80% cached) / 800 output	$1.24	$1.61	1.3×
Summarize a long report60,000 input / 1,200 output tokens	$6.60	$8.58	1.3×
Classify a ticket400 input / 10 output tokens	$0.045	$0.0585	1.3×

Standard pricing, no batch discount. Claude 4.7+ costs include its ~30% higher token count for the same text. Price is only one factor; quality, speed, context length and rate limits matter too, so test both on your own prompts.

Run your own numbers in the calculator →

More comparisons

More free tools from us

CleanStatement →

Convert bank and credit card statement PDFs to Excel or CSV. Runs in your browser; the file is never uploaded.

FillDesk →

Answer vendor security questionnaires from your own answer library and export to Word or Excel.