Haiku 5.5 vs Luna 6:
one prompt, both models.
Both stream side by side, with first-token latency, total time, tokens and what the run cost at list price.
2 free runs per day
Haiku 5.5anthropic/claude-haiku-5.5
- First token
- —
- Total
- —
- Tokens in / out
- —
- This run
- —
Luna 6openai/gpt-6-luna
- First token
- —
- Total
- —
- Tokens in / out
- —
- This run
- —
Free runs used up for today. Keep testing with your own OpenRouter key, or pick up a credit pack.
No daily limit with your own OpenRouter keyStays in your browser. Prompts and keys are never stored.
Default reasoning settings, up to ~2K input and 1.5K output tokens per model. Costs are list prices from the token counts each provider returns.
A fair test
Use a prompt from your real workload and run it a few times. One run is a sample, not a benchmark.
Why costs differ
Both charge $0.10 / $0.50 per million tokens at this size, so a cost gap comes from how much each one writes.
Price a full workload
Thresholds, caching and batch are in the calculator.