A side-by-side comparison of Claude Opus 4.8 and GPT-5.6 Luna: input/output pricing, capabilities and available endpoints, served live from AI中转站. Both are reachable with the same API key.
| Item | Claude Opus 4.8 | GPT-5.6 Luna |
|---|---|---|
| Input (per 1M tokens) | $4.5 | $0.65 |
| Output (per 1M tokens) | $22.5 | $3.9 |
| Cache read (per 1M tokens) | $0.45 | $0.065 |
| Cache write 5m (per 1M tokens) | $5.625 | $0.8125 |
| Item | Claude Opus 4.8 | GPT-5.6 Luna |
|---|---|---|
| Context window | 1,000,000 | 922,000 |
| Max output | 128,000 | 128,000 |
| Item | Claude Opus 4.8 | GPT-5.6 Luna |
|---|---|---|
| function_calling | Yes | Yes |
| prompt_caching | Yes | Yes |
| vision | Yes | Yes |
On input, GPT-5.6 Luna is cheaper ($4.5 vs $0.65 per 1M tokens). On output, GPT-5.6 Luna is cheaper ($22.5 vs $3.9 per 1M tokens). All prices are per million tokens in USD.
Claude Opus 4.8 does — Claude Opus 4.8 accepts 1,000,000 input tokens and GPT-5.6 Luna accepts 922,000.
both support function_calling, prompt_caching, vision.
Yes. Both are available on AI中转站 through one API key and the same OpenAI-compatible endpoint — switching means changing the model field from "claude-opus-4-8" to "gpt-5.6-luna", nothing else.
Both are available on AI中转站 under one API key — switching between them means changing the model field and nothing else, so you can use each where it fits rather than picking one.