Model
GPT-6 Luna evaluated on the hub's 2 benchmarks
Released on 22 September 2026, GPT-6 Luna (OpenAI) joins the hub's rankings: its place on the Business Index, its strongest and weakest business functions, its cost and its hallucination rate.
Report written by the hub from the published rankings: every sentence is computed, none is an opinion.
- 12th of 27 on the Business Index, at 98.5% (95% margin of error: ± 2.2 points).
- 1.2 index points below GPT-6 Astra, its predecessor at OpenAI, released on 4 September 2026. The gap is smaller than the margin of error: the two models cannot be told apart.
- Mean cost per test: $0.0006, the 4th cheapest of the 27 models with a known price.
- Mean hallucination rate: 0.0%, the share of items missing from the source for which it invented a value.
- It took all 2 of the hub's benchmarks.
Scores by business function
| Business function | Score |
|---|---|
| Finance | 96.9% |
| Accounting | 100.0% |
| Human resources | Not tested |
| Legal | Not tested |
| Sales | Not tested |
| Marketing | Not tested |
| Customer service | Not tested |
| Procurement & logistics | Not tested |
| IT | Not tested |
| Management | Not tested |
| Business Index | 98.5% |