NVIDIA: Llama 3.3 Nemotron Super 49B V1.5 wins on more metrics (4 of 7), but the right pick depends on what you optimize for — see the breakdown below.
| Metric | Nous: Hermes 4 405B | NVIDIA: Llama 3.3 Nemotron Super 49B V1.5 |
|---|---|---|
| Intelligence Index | 18.6✓ | 14.6 |
| Coding Index | 16.0✓ | 10.5 |
| GPQA Diamond | 73%✓ | 48% |
| Design Arena Elo | — | — |
| Speed (tokens/sec) | 35 | 44✓ |
| Latency | 345ms | 246ms✓ |
| Input price /M | $1.00 | $0.400✓ |
| Output price /M | $3.00 | $0.400✓ |
| Context window | 131K | 131K |
| Capabilities | ReasoningJSON | ReasoningToolsJSON |