~/system-one-bench / compare/baseten-mammoth-1-vs-together-ai-tev1-4b-experimental
Head to head
Mammoth 1 scores 82.9 and Tev1 4B Experimental 71.5 on the System One Index. Mammoth 1 is more accurate on 15 of 16 text datasets, Tev1 4B Experimental on 1. Same cases, same request, same scorer.
| Mammoth 1 | Tev1 4B Experimental | |
|---|---|---|
| System One Index | 82.9 | 71.5 |
| English | 82.5 | 66.4 |
| Arabic | 82.8 | 75.8 |
| Vision Index | โ | โ |
| Calibration error | 4.9% | 5.3% |
| Median latency | โ | 232 ms |
| Cost per 1,000 calls | $0.037 | $0.019 |
| Mammoth 1 | Tev1 4B Experimental | |
|---|---|---|
| AG News English | 92.0 | 91.5 |
| SST-2 English | 96.0 | 94.0 |
| SST-5 English | 59.0 | 43.0 |
| DAIR Emotion English | 59.0 | 56.5 |
| Banking77 English | 94.5 | 1.0 |
| BoolQ English | 92.0 | 91.0 |
| XNLI English English | 91.3 | 87.3 |
| Typed decisions English | 76.4 | 67.0 |
| XNLI French French | 86.0 | 82.7 |
| XNLI Arabic Arabic | 78.7 | 76.0 |
| MASSIVE scenario Arabic | 75.0 | 56.5 |
| MASSIVE intent Arabic | 91.0 | 70.0 |
| OSACT4 offensive Arabic | 91.5 | 84.5 |
| Tweet sentiment Arabic | 64.5 | 70.5 |
| MIRACL rerank Arabic | 89.0 | 85.5 |
| MIRACL pairwise Arabic | 90.0 | 87.5 |
Model pages: Mammoth 1 ยท Tev1 4B Experimental. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.