N
NVIDIA: Llama 3.3 Nemotron Super 49B V1.5
● Excellent8.1
/ 10
Llama-3.3-Nemotron-Super-49B-v1.5 is a 49B-parameter, English-centric reasoning/chat model derived from Meta’s Llama-3.3-70B-Instruct with a 128K context. It’s post-trained for agentic workflows (RAG, tool calling) via SFT across math, code, science, and...
Specifications
| Attribute | Value |
|---|---|
| Lab | Nvidia |
| Tags | Agentic Intelligent |
| Overall Score | 8.1/10 |
| Release Date | 2025-10 |
| Context Window | 131,072 tokens |
| Input Price / 1M | $0.40 |
| Output Price / 1M | $0.40 |
| Input Modalities | Text |
| Output Modalities | Text |
Strengths
- 49B parameters derived from Llama 3.3 70B with post-training
- Optimized for agentic workflows: RAG, tool calling, code
- 128K context window for complex multi-step tasks
- Competitive pricing at $0.40/$0.40 per 1M tokens
- Strong across math, code, science, and instruction following
Weaknesses
- Derivative model: not trained from scratch by NVIDIA
- Only moderately improved over base Llama 3.3 70B
- English-centric; limited multilingual performance
Best For
- Agentic applications requiring tool calling
- RAG pipelines and knowledge retrieval
- Enterprise chat and instruction following
- Code generation and technical tasks
Sources & Further Reading
Related Models
Scores are aggregated from public benchmarks (MMLU, HumanEval, MATH, GSM8K, LMSYS) and normalized to a 1–10 scale. Methodology →