# NVIDIA Nemotron 3.5 Bolt: 256K Context, 25GB Local Model

URL: https://3ammarketer.com/news/nemotron-3-5-bolt-local-vision-language-model
Published: 2026-10-24

> NVIDIA has released Nemotron 3.5 Bolt, a compact vision-language model for local coding, reasoning, and tool use featuring E4B and 74B-A9B architectures.

NVIDIA launched Nemotron 3.5 Bolt as a compact local vision language model built for local coding, reasoning, and tool use, requiring 25GB of storage and featuring a 256K context window for its latest build. The release introduces two primary deployment pathways, separating smaller desktop hardware from larger multi-accelerator configurations. [1]

## Understanding the Nemotron 3.5 Bolt architecture

![A screenshot of the Nemotron 3.5 Bolt model page on Ollama highlighting the model description and variants.](https://dygolixznokmekyvajbb.supabase.co/storage/v1/object/public/images/content/01f901ddf1e07b45358d0b54.png)

The 74B-A9B variant uses a hybrid structural design, combining a Mamba-2/Transformer mixture-of-experts backbone with a C-RADIO vision encoder. It maintains 74 billion total parameters while activating 9 billion parameters per token during processing, reducing active compute overhead while retaining full parameter capacity across text and image tasks. [1]

By incorporating the C-RADIO vision encoder directly into the Mamba-2 hybrid backbone, the architecture processes multimodal input feeds within its mixture-of-experts structure. For engineers evaluating [multimodal AI app architectures](/blog/multimodal-ai-app-costs-architecture), this design balances memory footprint and reasoning throughput during multimodal token generation. [1]

## Variant comparison: E4B vs 74B-A9B

Hardware compatibility sharply divides the Nemotron 3.5 Bolt family. The compact E4B model is post-trained from Gemma 4 E4B, packaging 8 billion total parameters with 4 billion effective parameters. In contrast, the larger 74B-A9B model targets server-grade inference and lacks support on Apple Silicon and Metal platforms. [1]


| Variant | Total Parameters | Active / Effective Parameters | Context Window | Storage Footprint | Supported Platforms |
| --- | --- | --- | --- | --- | --- |
| nemotron-3.5-bolt:e4b | 8B | 4B effective | 128K | 7.2GB - 7.9GB | Windows, Linux, Apple Silicon |
| nemotron-3.5-bolt:74b-a9b | 74B | 9B active per token | 256K | 25GB (base latest) | Windows, Linux (No Apple Silicon / Metal) |

Context window capacity of Nemotron 3.5 Bolt variants in tokens. [1]

- E4B: 128 K
- 74B-A9B: 256 K

Comparison of parameter counts and context window size between Nemotron 3.5 Bolt variants. [1]

- E4B Total Params: 8
- E4B Effective Params: 4
- 74B-A9B Total Params: 74
- 74B-A9B Active Params: 9

Teams deploying on consumer workstations can review [local AI model hardware requirements](/blog/local-ai-models-guide) to size memory channels, as the E4B build operates within 7.2GB to 7.9GB of storage and handles a 128K context window across Windows, Linux, and Apple Silicon environments. [1]

## Optimization and quantization techniques

NVIDIA created Bolt by compressing Nemotron 3.5 Super through further training and quantization-aware distillation. The published GGUF implementation applies IQ2_XXS quantization specifically to expert weights while preserving precision by assigning NVFP4 and Q8_0 formats across the remaining network layers. [1]

All Nemotron 3.5 Bolt releases are governed under the OpenMDW License Agreement, version 1.1. According to 3AM Marketer tracking data on the [open model release tracker](/data/open-models), NVIDIA has standardized on the openmdw-1.1 license across its recent open model releases. [1][3]

## Implementing local tool use and reasoning

Nemotron 3.5 Bolt natively supports multi-turn tool calling, permitting developers to execute functions and ingest operational responses across prolonged execution loops. When building with [local runtimes like Ollama](/blog/ollama-ai-local-runtime-specs-api-costs), the model calls functions and incorporates their results directly across multi-turn tasks via the documented runtime API. [1][2]

Developers can also toggle a thinking feature on or off to dictate model behavior. Enabling thinking prompts the model to execute explicit reasoning steps before returning its final answer. [1]

Limit: The public Ollama repository documents that Apple Silicon and Metal environments cannot run the 74B-A9B variant, restricting local Mac deployments entirely to the E4B 8-billion-parameter build.

## Sources

1. [nemotron-3.5-bolt](https://ollama.com/library/nemotron-3.5-bolt)
2. [ollama/docs/api.md at main · ollama/ollama · GitHub](https://github.com/ollama/ollama/blob/main/docs/api.md)
3. [Open model release tracker (3AM Marketer data set)](https://3ammarketer.com/data/open-models)