Overview
Both names are officially documented public models. Compare the API configurations you intend to use rather than treating product names as a guarantee of equivalent behavior.
Feature Comparison
| Official API feature | Claude Haiku 5.5 | GPT-6 Luna |
|---|---|---|
| Model ID | claude-haiku-5-5 | gpt-6-luna |
| Input / output | Text and image / text | Text and image / text |
| Reasoning | Adaptive thinking | Configurable reasoning effort |
| Tool workflow | Claude API tools | Responses API function calling |
Pricing Comparison
Standard uncached USD per million tokens, written as input / output. Haiku changes tiers above 100K input tokens per request; Luna applies multipliers above 272K. The last Luna cell is arithmetic derived from its official 2× input and 1.5× output multipliers. These are not monthly volume thresholds.
| Input length per request | Haiku 5.5 USD / million | GPT-6 Luna USD / million |
|---|---|---|
| Up to 100K | $0.10 / $0.50 | $0.10 / $0.50 |
| Above 100K through 272K | $0.50 / $2.50 | $0.10 / $0.50 |
| Above 272K | $0.50 / $2.50 | $0.20 / $0.75 (calculated) |
Coding and Reasoning
Anthropic publishes comparisons against GPT-6 Luna, but that is developer-reported evidence. Our linked benchmark guide preserves attribution and missing values. We have not run a matched head-to-head test, so no universal performance winner is claimed.
Context Windows
Official API limits: Haiku 5.5 lists a 1M-token context window and 128K maximum output; GPT-6 Luna lists 1,050,000 context tokens and 128,000 maximum output tokens. Check endpoint constraints and billed long-context tiers before using the maximum.
Developer Use Cases
- Evaluate high-volume extraction and classification with a labeled dataset.
- Test coding and tool workflows with the same permissions and completion criteria.
- Measure the cost of long-context conversations separately from short requests.
Which One Should You Choose?
Start with the provider already integrated into your application, then compare correctness, tool reliability, latency and actual billed usage on a representative task set. Similar short-request list prices do not imply identical total cost. Unknown parameter counts are not a reason to invent a hardware comparison.
Frequently asked questions
Is GPT-6 Luna an official public model name?
Yes. OpenAI’s announcement and API documentation use this name and the gpt-6-luna identifier; both are linked here.
Which model is better for coding?
Published results are useful context, but we have not run a controlled comparison. Evaluate the same repository tasks, tools and reasoning budgets before choosing.
Go straight to the source
Reviewed Oct 8, 2026. Official statements, third-party results and arithmetic estimates are attributed separately. Unknowns reflect the sources reviewed on this date. No inference benchmarks were run by this site.