| Claude Haiku 4.5 | GPT-5.6 Luna | |
|---|---|---|
| Input, per 1M tokens | $1.00 | $0.20 |
| Output, per 1M tokens | $5.00 | $1.20 |
| Context window | 200K | 1.05M |
| Classification or routing (600 tokens in, 10 out), per 1,000 | $0.65 | $0.132 |
| Chat reply (2,000 in, 400 out), per 1,000 | $4.00 | $0.88 |
| Retrieval-augmented answer (8,000 in, 500 out), per 1,000 | $10.50 | $2.20 |
| Confidence signal | no | no |
| Tool calling | yes | yes |
| BANKING77, correct of 400 | 301 | not measured |
Which one should carry your traffic?
If GPT-5.6 Luna answers your requests as well as Claude Haiku 4.5, moving that traffic saves 78% on chat-shaped requests. Averages cannot tell you whether it does for your requests. GPT-5.6 Luna does not expose a confidence signal through OpenRouter, so Gimbix would test it as a full replacement rather than a first answer.
Claude Haiku 4.5 profile · GPT-5.6 Luna profile · Measure it on your traffic