Model comparison
Claude Haiku 4.5 vs GPT-5.6 Luna
GPT-5.6 Luna is the cheaper of the two on input tokens, by a factor of 5.0. Claude Haiku 4.5 lists $1 in and $5 out per million tokens with a 200K context window; GPT-5.6 Luna lists $0.20 and $1.20 with 1M.
| Claude Haiku 4.5 | GPT-5.6 Luna | |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Input / 1M tokens | $1 | $0.20 |
| Output / 1M tokens | $5 | $1.20 |
| Context window | 200K | 1M |
| Max output | 64K | - |
| Modality | Multimodal | Multimodal |
| Open weights | No | No |
| Released | 2025-10-01 | 2026 |
| Source | Provider documentation | Secondary, verify |
| Illustrative month* | $60 | $13 |
* 100 calls a day, about 10,000 tokens in and 2,000 out each, at list price, before caching, batching or volume discounts. Illustrative only. Pricing and context data verified 2026-08-07; "Secondary" figures come from pricing aggregators and trade press, so check the provider before relying on them. No benchmark scores: see methodology.
Claude Haiku 4.5
Anthropic's fast, cheap tier. The right choice for classification, extraction and the reading-heavy legwork in a multi-agent setup where a frontier model would be waste.
Claude Haiku 4.5 in full →GPT-5.6 Luna
The cheap tier of the GPT-5.6 family. At $0.20 per million input tokens it is priced to compete with the open-weight hosted models rather than with frontier tiers.
GPT-5.6 Luna in full →Other comparisons