Llama 3.1 Nemotron 70B Instruct vs Sonar Reasoning Pro
NVIDIA vs Perplexity — benchmarks, pricing, and capabilities side by side.
- •Sonar Reasoning Pro has the higher intelligence index (95.7 vs 43.7)
- •Llama 3.1 Nemotron 70B Instruct is cheaper ($1.20 vs $2.00 per 1M input)
| Llama 3.1 Nemotron 70B Instruct | Sonar Reasoning Pro | |
|---|---|---|
| Intelligence index | 43.7 | 95.7 |
| Developer | NVIDIA | Perplexity |
| Type | LLM | Multimodal |
| Access | Open weights | API only |
| Context window | — | 128,000 tokens |
| Input price | $1.20 / 1M | $2.00 / 1M |
| Output price | $1.20 / 1M | $8.00 / 1M |
| Speed | 292 tok/s | — |
| Released | October 1, 2024 | March 7, 2025 |
| Parameters | 70000000000 | — |
| Input modalities | — | Text, Image |
| Output modalities | — | Text |
Shared benchmarks
Llama 3.1 Nemotron 70B Instruct
Sonar Reasoning Pro
MATH-500
73.3
95.7