back to local model checker

Nemotron 3.5 Bolt 74B-A9B

MoEVision

NVIDIA · 74B (9B) · Mixture of Experts

Run Nemotron 3.5 Bolt 74B-A9B locally. Check hardware compatibility, VRAM requirements, and performance.

OpenMDW 1.1
Downloads

0

Likes

0

Released

2026-10

Context

256K tokens

Use cases

chatvisionreasoningcode

Your hardware

16 GB Reference Desktop

Quant
Q4_K_M
Memory
38.4 GB
Speed
~1 tok/s
Fit
Barely runs

Quantization options

What is quantization?
QuantBitsVRAMDiskQualityStatus
Q2_K224.2 GB22.6 GBlowBarely runs · ~3 t/s
Q3_K_M333.7 GB31.7 GBmoderateBarely runs · ~1 t/s
Q4_K_M438.4 GB36.2 GBgoodBarely runs · ~1 t/s
Q5_K_M547.9 GB45.2 GBgoodToo heavy · ~1 t/s
Q6_K657.4 GB54.3 GBexcellentToo heavy · ~1 t/s
Q8_0876.3 GB72.4 GBexcellentToo heavy · ~1 t/s
F1616152.1 GB144.7 GBlosslessToo heavy · ~1 t/s

About this model

Nemotron 3.5 Bolt 74B-A9B (Nemotron 3.5 bolt 74b) is a 74B sparse mixture-of-experts (MoE) model activating 9B per token from NVIDIA in the NVIDIA family, designed primarily for chat, vision, reasoning, code workflows. Supports up to 256K context tokens; start at 4K–8K with Q8_0 KV cache on tight GPUs.

Hardware planning baselines start at 24 GB minimum memory and 38 GB recommended memory. At Q4_K_M, weights plus baseline runtime overhead use roughly 38.4 GB of RAM/VRAM and 36.2 GB of disk space. At 74B, Nemotron 3.5 Bolt 74B-A9B belongs firmly in the workstation and multi-GPU tier. Running it comfortably requires 38 GB of total memory across dual RTX GPUs, high-memory Apple Silicon (Mac Studio Ultra/Max), or MoE hybrid CPU-RAM offloading—mobile devices and standard 16 GB laptops are too constrained for practical speeds.

Deploy Nemotron 3.5 Bolt 74B-A9B locally via llama.cpp with Flash Attention enabled. Under OpenMDW 1.1, review official weight cards before commercial redistribution.

Highlights

  • Local Fit Baseline24 GB minimum and 38 GB recommended RAM/VRAM.
  • Primary QuantQ4_K_M targets 38.4 GB memory and 36.2 GB disk.
  • Context & Capabilities256K max context · Sparse MoE.

System requirements by device

Memory planning targets 24 GB minimum RAM/VRAM and 38 GB recommended memory. The baseline quant is Q4_K_M, estimated at 38.4 GB memory and 36.2 GB disk. Actual runtime usage shifts with context length, KV cache precision, and OS background overhead.

DeviceNotes
Mac (Apple silicon)Requires a 48 GB–192 GB Mac Studio or MacBook Pro Max/Ultra with enough unified wired budget for 38.4 GB.
Windows / Linux GPUUse dual/multi-GPU CUDA splitting (--tensor-split) or MoE CPU expert offloading (-ot) paired with 32–128 GB DDR5 system RAM.
Phones & Compact Edge BoardsNot practical on phones (24 GB min RAM exceeds Android LMK and iOS Jetsam limits). Pick a 2B–4B model on mobile.

Start with a short session and leave memory for the operating system. A file fitting on disk does not guarantee that it will load in memory.

Estimated memory by quantization

These are rough estimates, not tested minimum requirements. Check the actual file and runtime documentation; context, offloading and operating-system overhead can change the memory needed.

Quant TierEstimated memoryDisk SizeQuality
Q2_K24.2 GB22.6 GBlow
Q3_K_M33.7 GB31.7 GBmoderate
Q4_K_M38.4 GB36.2 GBgood
Q5_K_M47.9 GB45.2 GBgood
Q6_K57.4 GB54.3 GBexcellent
Q8_076.3 GB72.4 GBexcellent
F16152.1 GB144.7 GBlossless

How to run Nemotron 3.5 Bolt 74B-A9B on your device

Start with a short prompt and the smallest quant that still gives the quality you need. A clean first run is more useful than loading a larger file with no memory headroom.

Windows

Run Nemotron 3.5 Bolt 74B-A9B on Multi-GPU or High-RAM Desktop

  1. Plan for 24 GB minimum combined memory and 38 GB recommended memory for Q4_K_M.
  2. Use llama.cpp `-ot ".ffn_.*_exps.=CPU"` or `--tensor-split` across multiple GPUs so attention and KV cache remain in fast VRAM.
  3. Enable Flash Attention (`-fa on`) and Q8_0 KV cache (`-ctk q8_0 -ctv q8_0`) to prevent context buffers from exceeding GPU headroom.
  4. Avoid attempting to run workstation-scale weights on smartphones or 8 GB laptops; use a 2B–8B checkpoint on portable devices.
Device how-to guide
Mac

Run Nemotron 3.5 Bolt 74B-A9B on High-Memory Apple Silicon Mac

  1. Verify you are on a 48 GB–192 GB Mac Studio or MacBook Pro Max/Ultra capable of holding 38.4 GB in wired unified memory.
  2. Launch via `llama-server` or LM Studio with MLX/Metal enabled and set initial context conservatively to 4K–8K.
  3. Monitor Activity Monitor Memory Pressure to confirm zero SSD swap thrashing during token generation.
  4. Step down to a lighter quantization (e.g., Q3_K_M or IQ4_XS) if total memory utilization exceeds 75%.
Device how-to guide

Nemotron 3.5 Bolt 74B-A9B vs alternatives

ModelParamsContextMin RAMLicense
Nemotron 3.5 Bolt 74B-A9B74B (9B)256K24 GBOpenMDW 1.1
Qwen 2.5 72B72B128K40.2 GBQwen
Llama 3.3 70B70B128K39.1 GBLlama 3.3 Community
Qwen 3 Next 80B-A3B80B (3B active)256K44.7 GBApache 2.0
Qwen 3 Coder Next 80B-A3B80B (3B active)256K44.7 GBApache 2.0
HunyuanImage 3.080B (13B active)4K44.7 GBTencent Hunyuan Community

Use this table to choose the smallest nearby model that still matches the job. A related model with lower RAM can be the better local choice when the current page fits only at the edge of available memory.

Frequently asked questions

What is Nemotron 3.5 Bolt 74B-A9B?
Nemotron 3.5 Bolt 74B-A9B is a 74B sparse mixture-of-experts model (9B active per token) from NVIDIA, supporting up to 256K tokens with 24 GB minimum RAM and 38 GB recommended RAM.
What is the best way to run Nemotron 3.5 bolt 74b locally?
Choose the smallest quant that matches the task, test one short prompt, then compare RAM, context, and nearby models before making Nemotron 3.5 Bolt 74B-A9B your default local model.
How much VRAM or RAM does Nemotron 3.5 Bolt 74B-A9B require?
At Q4_K_M, Nemotron 3.5 Bolt 74B-A9B occupies roughly 38.4 GB of active memory and 36.2 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (38 GB recommended).
What is the best way to run Nemotron 3.5 Bolt 74B-A9B locally?
Run Nemotron 3.5 Bolt 74B-A9B in llama.cpp starting with Q4_K_M. Keep initial context at 4K–8K tokens and confirm 100% GPU/Metal offload before scaling context length.
Can Nemotron 3.5 Bolt 74B-A9B run on a consumer laptop or smartphone?
No smartphone or 8–16 GB laptop can hold Nemotron 3.5 Bolt 74B-A9B cleanly in local memory. It targets multi-GPU NVIDIA workstations, high-RAM Mac Studios, or MoE CPU-RAM hybrid rigs with at least 24 GB total memory.
How does Nemotron 3.5 Bolt 74B-A9B compare to Qwen 2.5 72B?
Qwen 2.5 72B (72B, 40.2 GB min RAM) sits closest in size. Compare active parameters, context ceiling (256K vs 128K), and license terms (OpenMDW 1.1) before switching.