Nvidia to release Nemotron Bolt 3.5 74B A9B agent model

A 74B MoE squeezed into 24GB could make capable agent models run on consumer GPUs.

Nvidia's Nemotron Bolt 3.5 is a 74B-parameter, 9B-active MoE model using 2-bit quantization to fit in 24GB, aimed at fast agent workloads; release reported for Oct. 15.

Name

Nemotron Bolt 3.5 74B A9B

Params

74B total, 9B active

Read it at Reddit →

Coverage: 2 sources, in order

1. 14 Oct 14:42 Reddit: NVIDIA Nemotron Bolt 3.5 74B A9B releases Oct. 15th
2. 15 Oct 02:51 OrcaRouter: Nemotron 3.5 Bolt 74B Release Date: 74B Agent in 24GB

← Front page