| Result | Rank 1GPT-6 Luna None | Rank 2GPT-6 Luna Low | Rank 3GPT-6 Sol Low | Rank 4Claude Haiku 5.5 Low | Rank 5Claude Sonnet 5.5 Medium | Rank 6Claude Opus 5.5 Medium |
|---|---|---|---|---|---|---|
| Combined score | 88.83 | 88.80 | 88.69 | 74.98 | 73.17 | 67.78 |
| Judge | ||||||
| Judge score /100 | 88.50 | 88.50 | 88.50 | 75.00 | 73.00 | 67.50 |
| Correctness /10 | 9 | 9 | 9 | 8 | 8 | 7 |
| Coverage /10 | 9 | 9 | 9 | 6 | 6 | 6 |
| Evidence /10 | 9 | 9 | 9 | 8 | 7 | 7 |
| Actionability /10 | 8 | 8 | 8 | 8 | 8 | 7 |
| Time | 3.3 s | 3.65 s | 5.17 s | 20.37 s | 18.7 s | 22.42 s |
| Estimated USD | $0.0003328 | $0.0003393 | $0.007114 | N/A | N/A | N/A |
Supplemental estimates use reviewed exact-route API-equivalent list-rate scenarios with cache ignored; recorded estimates retain their frozen rates. Cost remains excluded from this battle score. These estimates are not Codex subscription charges or billed USD.
Historical Claude CLI cache counters were not retained. Complete input and complete reference cost are unavailable.
Judge score uses recorded rubric metrics; final rank uses the combined score.
Component points and token usage
| Result | Rank 1 GPT-6 Luna None | Rank 2 GPT-6 Luna Low | Rank 3 GPT-6 Sol Low | Rank 4 Claude Haiku 5.5 Low | Rank 5 Claude Sonnet 5.5 Medium | Rank 6 Claude Opus 5.5 Medium |
|---|---|---|---|---|---|---|
| Quality points | 83.8421 | 83.8421 | 83.8421 | 71.0526 | 69.1579 | 63.9474 |
| Test points | 0 | 0 | 0 | 0 | 0 | 0 |
| Speed points | 4.9887 | 4.9613 | 4.846 | 3.9294 | 4.0126 | 3.8314 |
| Cost points | 0 | 0 | 0 | 0 | 0 | 0 |
| Total input incl cache | N/A | N/A | 2,352 | N/A | N/A | N/A |
| Cache read (subset of input) | N/A | N/A | 0 | N/A | N/A | N/A |
| Cache write (subset of input) | N/A | N/A | 0 | N/A | N/A | N/A |
| Total output incl reasoning | N/A | N/A | 241 | N/A | N/A | N/A |
| Reasoning (subset of output) | N/A | N/A | N/A | N/A | N/A | N/A |
Swipe, scroll or use Previous and More to compare configurations. Row labels stay visible.
Advanced results and scoring details
engineering · 2026-10-13T05:12:38.543Z · Prospective evaluation. Equal transport and native system conditions are unverified.
Applied weights: quality 94.7368%, tests 0%, speed 5.2632%, cost 0%.
Cost was omitted for every configuration in this cohort. Active weights were normalized uniformly. Displayed scores are rounded; pairwise outcomes use exact recorded fractions. Reference costs are estimates, not billed charges. Missing components follow the recorded assessment.
Pairwise results
- Claude Opus 5.5 Medium vs GPT-6 Luna None: Second configuration wins
- Claude Opus 5.5 Medium vs Claude Haiku 5.5 Low: Second configuration wins
- Claude Opus 5.5 Medium vs GPT-6 Luna Low: Second configuration wins
- Claude Opus 5.5 Medium vs GPT-6 Sol Low: Second configuration wins
- Claude Opus 5.5 Medium vs Claude Sonnet 5.5 Medium: Second configuration wins
- GPT-6 Luna None vs Claude Haiku 5.5 Low: First configuration wins
- GPT-6 Luna None vs GPT-6 Luna Low: First configuration wins
- GPT-6 Luna None vs GPT-6 Sol Low: First configuration wins
- GPT-6 Luna None vs Claude Sonnet 5.5 Medium: First configuration wins
- Claude Haiku 5.5 Low vs GPT-6 Luna Low: Second configuration wins
- Claude Haiku 5.5 Low vs GPT-6 Sol Low: Second configuration wins
- Claude Haiku 5.5 Low vs Claude Sonnet 5.5 Medium: First configuration wins
- GPT-6 Luna Low vs GPT-6 Sol Low: First configuration wins
- GPT-6 Luna Low vs Claude Sonnet 5.5 Medium: First configuration wins
- GPT-6 Sol Low vs Claude Sonnet 5.5 Medium: First configuration wins