Nemotron Bolt 3.5 74B A9B
An open-source Mixture-of-Experts (MoE) large language model released by NVIDIA. It has 74B total parameters with roughly 9B activated per token, and is compressed to approximately 24GB of VRAM through 2-bit quantization-aware training, targeting local deployment on consumer-grade GPUs.
Core Facts
Timeline (last 90 days)
Messages circulating in the Reddit community claim that NVIDIA Nemotron Bolt 3.5 74B A9B will be released on October 15.
Nemotron Bolt 3.5 74B A9B is a mixture-of-experts (MoE) model with about 74B total parameters, activating only about 9B parameters per token.
Nemotron Bolt 3.5 74B A9B uses 2-bit quantization combined with quantization-aware training (QAT) to compress the entire model to about 24GB of VRAM usage.
NVIDIA claims that Nemotron Bolt 3.5 is faster than most popular dense models while being smarter than mainstream MoE models with the same memory footprint.
All Facts (4)
Nemotron Bolt 3.5 74B A9B is a mixture-of-experts (MoE) model with about 74B total parameters, activating only about 9B parameters per token.
75%UnverifiedNemotron Bolt 3.5 74B A9B uses 2-bit quantization combined with quantization-aware training (QAT) to compress the entire model to about 24GB of VRAM usage.
50%UnverifiedNVIDIA claims that Nemotron Bolt 3.5 is faster than most popular dense models while being smarter than mainstream MoE models with the same memory footprint.
50%UnverifiedMessages circulating in the Reddit community claim that NVIDIA Nemotron Bolt 3.5 74B A9B will be released on October 15.
50%