Head to head
d1-omni-600M scores 65.2 and GPT-6 Luna Decisions 76.0 on the System One Index. d1-omni-600M is more accurate on 4 of 16 text datasets, GPT-6 Luna Decisions on 11, and 1 are tied. Same cases, same request, same scorer.
| d1-omni-600M | GPT-6 Luna Decisions | |
|---|---|---|
| System One Index | 65.2 | 76.0 |
| English | 75.5 | 75.2 |
| Arabic | 53.4 | 78.8 |
| Vision Index | – | 81.6 |
| Calibration error | 9.8% | 9.8% |
| Median latency | – | 215 ms |
| Cost per 1,000 calls | – | $0.040 |
| d1-omni-600M | GPT-6 Luna Decisions | |
|---|---|---|
| AG News English | 87.5 | 87.5 |
| SST-2 English | 86.5 | 93.5 |
| SST-5 English | 47.0 | 61.5 |
| DAIR Emotion English | 85.0 | 58.5 |
| Banking77 English | 84.5 | 77.5 |
| BoolQ English | 74.0 | 85.5 |
| XNLI English English | 82.0 | 66.7 |
| Typed decisions English | 57.5 | 71.0 |
| XNLI French French | 65.3 | 62.0 |
| XNLI Arabic Arabic | 52.7 | 54.0 |
| MASSIVE scenario Arabic | 37.5 | 66.0 |
| MASSIVE intent Arabic | 52.0 | 84.5 |
| OSACT4 offensive Arabic | 77.5 | 93.5 |
| Tweet sentiment Arabic | 32.5 | 80.0 |
| MIRACL rerank Arabic | 55.0 | 86.0 |
| MIRACL pairwise Arabic | 66.5 | 87.5 |
| d1-omni-600M | GPT-6 Luna Decisions | |
|---|---|---|
| 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: d1-omni-600M · GPT-6 Luna Decisions. Latency for API models is the median per call through OpenRouter, timed from one machine.