GPT-6 Luna vs Claude Haiku 5.5
At 6,000 chat requests per month (1,500 in / 400 out), the two are within a few percent: $2.10 versus $2.10.
| Spec | GPT-6 Luna | Claude Haiku 5.5 |
|---|---|---|
| Vendor | OpenAI | Anthropic |
| Input / 1M tokens | $0.100 | $0.100 |
| Output / 1M tokens | $0.500 | $0.500 |
| Cached input / 1M | $0.010 | $0.010 |
| Context window | 1.1M tokens | 1M tokens |
| Max output | 128K tokens | 128K tokens |
| Input modalities | text, image | text, image |
| Released | — | — |
| Cost per 1M cached reads | $0.010 | $0.010 |
Published list prices, last verified Oct 15, 2026. Source links are on each model's page.
Cost at three workloads
| Workload | GPT-6 Luna | Claude Haiku 5.5 | Difference |
|---|---|---|---|
| Customer-support reply (1,500 in / 400 out) | $0.0004 | $0.0004 | — |
| Document summary (20,000 in / 1,000 out) | $0.0025 | $0.0025 | — |
| Code review pass (8,000 in / 1,500 out) | $0.0016 | $0.0016 | — |
Which one fits your workload
- Long context: GPT-6 Luna accepts more — 1.1M versus 1M tokens. If your retrieved context regularly exceeds 1M, only one of these will hold it.
- Long answers: output is priced the same per token, so the difference is purely how many tokens you generate.
- High-volume, short answers: input is priced identically.
- Caching: at least one of these vendors publishes a discounted cached-input rate; if you resend the same system prompt and context, model the cache-hit rate or you will overestimate cost.
This comparison uses published price and specification data only. It does not rank output quality — run your own evaluation on your own prompts before switching a production default.