Head to head
GPT-6 Luna Decisions scores 76.0 and d3 edge 65.9 on the System One Index. GPT-6 Luna Decisions is more accurate on 12 of 16 text datasets, d3 edge on 4. Same cases, same request, same scorer.
| GPT-6 Luna Decisions | d3 edge | |
|---|---|---|
| System One Index | 76.0 | 65.9 |
| English | 75.2 | 69.1 |
| Arabic | 78.8 | 62.2 |
| Vision Index | 81.6 | – |
| Calibration error | 9.8% | 18.7% |
| Median latency | 215 ms | – |
| Cost per 1,000 calls | $0.040 | – |
| GPT-6 Luna Decisions | d3 edge | |
|---|---|---|
| AG News English | 87.5 | 80.0 |
| SST-2 English | 93.5 | 80.5 |
| SST-5 English | 61.5 | 39.5 |
| DAIR Emotion English | 58.5 | 57.0 |
| Banking77 English | 77.5 | 90.0 |
| BoolQ English | 85.5 | 79.5 |
| XNLI English English | 66.7 | 84.0 |
| Typed decisions English | 71.0 | 42.3 |
| XNLI French French | 62.0 | 66.0 |
| XNLI Arabic Arabic | 54.0 | 59.3 |
| MASSIVE scenario Arabic | 66.0 | 41.0 |
| MASSIVE intent Arabic | 84.5 | 55.0 |
| OSACT4 offensive Arabic | 93.5 | 72.5 |
| Tweet sentiment Arabic | 80.0 | 61.5 |
| MIRACL rerank Arabic | 86.0 | 70.0 |
| MIRACL pairwise Arabic | 87.5 | 76.0 |
| GPT-6 Luna Decisions | d3 edge | |
|---|---|---|
| SST-2, as an image | 88.5 | – |
| Arabic sentiment, as an image | 67.0 | – |
| Oxford-IIIT Pet | 87.0 | – |
| VQAv2 yes/no | 80.5 | – |
| CountBench | 85.0 | – |
Model pages: GPT-6 Luna Decisions · d3 edge. Latency for API models is the median per call through OpenRouter, timed from one machine.