GPT-6 Luna vs Claude Haiku 5.5
Compare with other modelsGPT-6 Luna and Haiku 5.5 have similar quality scores, at 87 and 84. The bigger difference is the share of conversations with no unsupported claims: 90% for Luna and 62% for Haiku. Luna also costs about half as much per conversation.
Haiku replies faster, with a median of 2.6 seconds against Luna's 4.8, and it scores higher on grounding. Start with Luna if unsupported claims are the main concern, and try Haiku if reply speed matters more.
| Metric | OpenAI Sep 23, 2026 | Anthropic Oct 7, 2026 |
|---|---|---|
| Overview | ||
| Quality score | 87 #1 of 17 | 84 #4 of 17 |
| Median reply | 4.8s Served by OpenAI | 2.6s Best among selected models Served by Claude Platform on AWS |
| Cost per conversation | $0.0040 Best among selected models | $0.0076 |
| Quality scores | ||
| Tool use | 83 | 86 Best among selected models |
| Task completion | 85 Best among selected models | 84 |
| Context retention | 85 Best among selected models | 81 |
| Grounding | 87 | 96 Best among selected models |
| Safety | 93 | 94 Best among selected models |
| Hallucinations % of conversations with no unsupported claims detected | 90% Best among selected models | 62% |
| Running stats | ||
| p95 reply | 14.0s | 7.2s Best among selected models |
| Consistency % of repeat runs ended the same way | 77% | 87% Best among selected models |
| Output per turn | 267 tokens, median | 338 tokens, median |
Reply times cover the model call only. Cost estimates use token usage at list prices.
This whole report is one Voxli workspace: simulated customers, assertion checks, and a frozen, versioned test set that reruns when new models ship.
Test set v3f-2026-09 · edition 2026-09-08