Nemotron 3.5 Bolt
Announced on nvidia.com
Benchmark results
| Benchmark | Score | Source check |
|---|---|---|
| Humanity's Last ExamNo tools · BF16 weights | 28.73% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| 24GB build with 2-bit (IQ2) weights | 25.30% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| GPQA DiamondBF16 weights | 88.83% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| 24GB build with 2-bit (IQ2) weights | 86.43% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| BrowseCompBF16 weights; Exa web search | 70.00% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| 24GB build with 2-bit (IQ2) weights; Exa web search | 66.75% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Terminal-Bench 2.1BF16 weights | 53.93% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| 24GB build with 2-bit (IQ2) weights | 50.98% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| MCP AtlasBF16 weights; 500 public tasks; reasoning on | 55.60% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| 24GB build with 2-bit (IQ2) weights; 500 public tasks; reasoning on | 52.20% | CorroboratedThe same number is printed in 1 other lab document: 1 |
Sources
Reported by NVIDIA
- Nemotron 3.5 Bolt 74B-A9B 24GB: A Hybrid Vision-Language Model for Local AIresearch.nvidia.com · 15 Oct 2026