Claude Haiku 4.5 vs GPT-5.6 Luna: API pricing comparison
Anthropic Claude Haiku 4.5 vs OpenAI GPT-5.6 Luna: token prices, batch and cache rates, Arena ELO, and cost per request for 8 business use cases. Prices in USD per 1 million tokens. Data snapshot: 2026-11-12.
Claude Haiku 4.5 vs GPT-5.6 Luna: the short version
At the prices we quote, GPT-5.6 Luna costs less than Claude Haiku 4.5 on input and on output: $0.20 against $1.00 per 1M input tokens (80% less), and $1.20 against $5.00 per 1M output tokens (76% less). With the Support Ticket profile (1,500 input and 500 output tokens per request), 1,000 requests cost $4.00 on Claude Haiku 4.5 and $0.90 on GPT-5.6 Luna at list price, 78% less on GPT-5.6 Luna. On Arena ELO, GPT-5.6 Luna scores 1431 and Claude Haiku 4.5 scores 1396, 35 points apart (scores as of 2026-11-08). The context window is 200K tokens on Claude Haiku 4.5 and 1M tokens on GPT-5.6 Luna. Both are recorded here as proprietary, with access through an API. Prices as quoted on 2026-11-12. Each price is the list price of one host, the one we quote for that model. A model sold by several hosts has other prices, listed on its model page.
Claude Haiku 4.5 pricing (Anthropic)
- Input: $1.00 per 1M tokens
- Output: $5.00 per 1M tokens
- Batch API (input/output): $0.50 / $2.50
- Prompt-cache read: $0.10
- Context window: 200K
- Arena ELO: 1396 (as of 2026-11-08)
- Full Claude Haiku 4.5 pricing page
GPT-5.6 Luna pricing (OpenAI)
- Input: $0.20 per 1M tokens
- Output: $1.20 per 1M tokens
- Batch API (input/output): $0.10 / $0.60
- Prompt-cache read: $0.02
- Context window: 1M
- Arena ELO: 1431 (as of 2026-11-08)
- Full GPT-5.6 Luna pricing page
Cost per 1,000 requests by business use case
List price vs after discounts (prompt caching on the cached input share, batch API for async workloads, only where the provider publishes those rates).
| Use case | Tokens (in/out) | Claude Haiku 4.5 list price | Claude Haiku 4.5 after discounts | GPT-5.6 Luna list price | GPT-5.6 Luna after discounts |
|---|---|---|---|---|---|
| Support Ticket | 1,500 / 500 | $4.00 | $3.24 | $0.90 | $0.75 |
| Knowledge Q&A | 2,000 / 800 | $6.00 | $4.81 | $1.36 | $1.12 |
| Meeting Summary | 10,000 / 1,200 | $16.00 | $7.75 | $3.44 | $1.67 |
| Marketing Content | 2,500 / 1,800 | $11.50 | $11.08 | $2.66 | $2.58 |
| Coding Task | 3,000 / 2,000 | $13.00 | $11.73 | $3.00 | $2.75 |
| Invoice Processing | 1,500 / 600 | $4.50 | $2.14 | $1.02 | $0.49 |
| Call Summary | 2,000 / 700 | $5.50 | $2.70 | $1.24 | $0.61 |
| Agent Workflow | 6,000 / 3,000 | $21.00 | $17.43 | $4.80 | $4.09 |
Common questions about Claude Haiku 4.5 vs GPT-5.6 Luna
- Which is cheaper, Claude Haiku 4.5 or GPT-5.6 Luna?
- At the prices we quote, GPT-5.6 Luna costs less than Claude Haiku 4.5 on input and on output: $0.20 against $1.00 per 1M input tokens (80% less), and $1.20 against $5.00 per 1M output tokens (76% less). Prices as quoted on 2026-11-12. Each price is the list price of one host, the one we quote for that model. A model sold by several hosts has other prices, listed on its model page.
- What do 1,000 Support Ticket requests cost on Claude Haiku 4.5 and on GPT-5.6 Luna?
- With the Support Ticket profile (1,500 input and 500 output tokens per request), 1,000 requests cost $4.00 on Claude Haiku 4.5 and $0.90 on GPT-5.6 Luna at list price, 78% less on GPT-5.6 Luna. Prices as quoted on 2026-11-12.
- How do Claude Haiku 4.5 and GPT-5.6 Luna compare on Arena score, context window and licence?
- On Arena ELO, GPT-5.6 Luna scores 1431 and Claude Haiku 4.5 scores 1396, 35 points apart (scores as of 2026-11-08). The context window is 200K tokens on Claude Haiku 4.5 and 1M tokens on GPT-5.6 Luna. Both are recorded here as proprietary, with access through an API.
Open this comparison in the interactive compare tool: add up to two more models, copy a shareable link, or export the table as CSV.
These are list prices: real cost also depends on caching, batching, negotiated discounts, marketplace agreements, and existing contracts. Get a FinOps review from OptimNow or compare all models.