Nemotron 3.5 Lightning 30B-A3B hardware requirements

What is Nemotron 3.5 Lightning 30B-A3B?

Nemotron 3.5 Lightning 30B-A3B is a general-purpose reasoning and chat model NVIDIA released in August 2026, also trained on coding and tool-calling data. It is a mixture-of-experts model that mixes Mamba-2 and attention layers, with about 30B parameters in total but about 3B active per token. Its thinking mode can be switched on or off in the chat template, and the OpenMDW-1.1 license allows commercial use.

Model card

About the model

Parameters: 31.58B total · 3B active (MoE)
Attention: Hybrid attention (only some layers use full attention)
Max context: 262,144 tokens
Released: 2026-08-01
Model card: Nemotron 3.5 Lightning 30B-A3B on Hugging Face

GGUF files

Tracked quantizations of Nemotron 3.5 Lightning 30B-A3B, smallest first
Quant	File size	Quality	Published by
Q2_K	18.91 GB	Heavy loss	Community GGUF · bartowski
IQ4_XS	18.92 GB	Good balance	Community GGUF · bartowski
Q4_K_Mbaseline	25.48 GB	Good balance	Community GGUF · bartowski
Q8_0	35.00 GB	Near-lossless	Community GGUF · bartowski

For this model, the Q2_K file is about the same size as the IQ4_XS file, which keeps more quality. Download IQ4_XS instead.

Memory needed by quant and context

Graphics card memory in GB: weights loaded on the GPU + f16 KV cache + compute buffer. Add about 0.6 GB if the GPU also drives your display on Windows. Part of the file stays off the graphics card: llama.cpp keeps the input embedding table in system RAM and does not use this model’s MTP (multi-token prediction) block yet, so it never loads it. A Mac counts the whole file, so it needs a little more.
Quant	4k	8k	16k	32k	64k	128k	256k
Q2_K	18.5	18.6	18.7	19.0	19.5	20.5	22.5
IQ4_XS	18.5	18.6	18.7	19.0	19.5	20.5	22.5
Q4_K_M	25.0	25.1	25.2	25.5	26.0	27.0	29.0
Q8_0	33.8	33.8	34.0	34.2	34.7	35.7	37.7

Which GPUs can run Nemotron 3.5 Lightning 30B-A3B?

How to read the verdicts

Runs great: Fully on the GPU at 20 tok/s or more
Runs well: Fully on the GPU at 8–20 tok/s; MoE experts in system RAM at 20 tok/s or more; or at least 90% on the GPU at 8 tok/s or more
Runs slowly: 2–8 tok/s; CPU-only; less than 90% on the GPU; or MoE experts in system RAM below 20 tok/s
Won't run: Does not fit, or under 2 tok/s

Sort

Filter this list by budget, type and context in the calculator

Scroll sideways to see every column.

56 GPUs, Macs and CPU setups at Q4_K_M with 8k context, sorted by verdict and speed
Hardware	Verdict	Speed	Quant	Memory	Runs as	Context	Notes	Price
GeForce RTX 5090	Runs great	est. 305.7 tok/s214.0–397.4theoretical estimate ±30%	Q4_K_M	25.7 / 32.0 GB	Full GPU	up to 256k	—	$4,200 street (as of 2026-08-09)
Apple M5 Ultra · 96 GB	Runs great	est. 168.6 tok/s118.0–219.2theoretical estimate ±30%	Q4_K_M	26.1 / 67.2 GB	Unified memory	up to 256k	—	—
Apple M5 Max (40-core GPU) · 48 GB	Runs great	est. 137.2 tok/s96.0–178.3theoretical estimate ±30%	Q4_K_M	26.1 / 33.6 GB	Unified memory	up to 256k	—	—
Apple M3 Ultra · 96 GB	Runs great	est. 126.5 tok/s88.5–164.4theoretical estimate ±30%	Q4_K_M	26.1 / 67.2 GB	Unified memory	up to 256k	—	—
Apple M4 Max (40-core GPU) · 48 GB	Runs great	est. 124.5 tok/s87.2–161.9theoretical estimate ±30%	Q4_K_M	26.1 / 33.6 GB	Unified memory	up to 256k	—	—
Apple M2 Ultra · 64 GB	Runs great	est. 124.2 tok/s87.0–161.5theoretical estimate ±30%	Q4_K_M	26.1 / 44.8 GB	Unified memory	up to 256k	—	—
Apple M5 Pro · 48 GB	Runs great	est. 75.6 tok/s52.9–98.3theoretical estimate ±30%	Q4_K_M	26.1 / 33.6 GB	Unified memory	up to 256k	—	—
NVIDIA DGX Spark · 64 GB	Runs great	est. 69.8 tok/s48.8–90.7theoretical estimate ±30%	Q4_K_M	26.1 / 48.2 GB	Unified memory	up to 256k	—	$3,999 launch MSRP (128 GB)
Apple M4 Pro · 48 GB	Runs great	est. 68.0 tok/s47.6–88.4theoretical estimate ±30%	Q4_K_M	26.1 / 33.6 GB	Unified memory	up to 256k	—	—
Ryzen AI Max+ 395 (Strix Halo) · 64 GB	Runs great	est. 65.9 tok/s46.1–85.7theoretical estimate ±30%	Q4_K_M	26.1 / 44.8 GB	Unified memory	up to 256k	—	$1,999 launch MSRP (128 GB)
GeForce RTX 3090 Ti	Runs well	est. 126.5 tok/s88.5–164.4theoretical estimate ±30%	Q4_K_M	22.7 / 24.0 GB + 4.0 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory Try IQ4_XS: Runs great	$1,999 launch MSRP
GeForce RTX 4090	Runs well	est. 126.5 tok/s88.5–164.4theoretical estimate ±30%	Q4_K_M	22.7 / 24.0 GB + 4.0 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory Try IQ4_XS: Runs great	$2,755 street (as of 2026-08-09)
GeForce RTX 3090	Runs well	est. 122.0 tok/s85.4–158.7theoretical estimate ±30%	Q4_K_M	22.7 / 24.0 GB + 4.0 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory Try IQ4_XS: Runs great	$1,050 street (as of 2026-08-09)
GeForce RTX 5090 Laptop	Runs well	est. 119.4 tok/s83.6–155.3theoretical estimate ±30%	Q4_K_M	22.7 / 24.0 GB + 4.0 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory Try IQ4_XS: Runs great	—
Radeon RX 7900 XTX	Runs well	est. 112.9 tok/s79.0–146.7theoretical estimate ±30%	Q4_K_M	22.7 / 24.0 GB + 4.0 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory Try IQ4_XS: Runs great	$999 launch MSRP
Radeon RX 7900 XT	Runs well	est. 74.5 tok/s52.1–96.8theoretical estimate ±30%	Q4_K_M	18.8 / 20.0 GB + 7.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory Try IQ4_XS: Runs great	$899 launch MSRP
GeForce RTX 5080	Runs well	est. 63.1 tok/s44.2–82.0theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$1,256 street (as of 2026-08-09)
GeForce RTX 5070 Ti	Runs well	est. 62.2 tok/s43.5–80.8theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$949 street (as of 2026-08-09)
GeForce RTX 5080 Laptop	Runs well	est. 62.2 tok/s43.5–80.8theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	—
GeForce RTX 4080 Super	Runs well	est. 59.3 tok/s41.5–77.1theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$999 launch MSRP
GeForce RTX 4080	Runs well	est. 58.9 tok/s41.2–76.5theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$1,199 launch MSRP
GeForce RTX 4070 Ti Super	Runs well	est. 57.8 tok/s40.5–75.2theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$799 launch MSRP
GeForce RTX 4090 Laptop	Runs well	est. 55.3 tok/s38.7–71.9theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	—
Radeon RX 9070	Runs well	est. 55.0 tok/s38.5–71.5theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$639 street (as of 2026-08-09)
Radeon RX 9070 XT	Runs well	est. 55.0 tok/s38.5–71.5theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$689 street (as of 2026-08-09)
Radeon RX 7800 XT	Runs well	est. 54.6 tok/s38.2–70.9theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$499 launch MSRP
GeForce RTX 5060 Ti 16GB	Runs well	est. 50.8 tok/s35.6–66.0theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$569 street (as of 2026-08-09)
GeForce RTX 3080 12GB	Runs well	est. 50.3 tok/s35.2–65.4theoretical estimate ±30%	Q4_K_M	10.8 / 12.0 GB + 15.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$799 launch MSRP
GeForce RTX 5070	Runs well	est. 47.5 tok/s33.3–61.8theoretical estimate ±30%	Q4_K_M	10.8 / 12.0 GB + 15.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$629 street (as of 2026-08-09)
GeForce RTX 5070 Ti Laptop	Runs well	est. 47.5 tok/s33.3–61.8theoretical estimate ±30%	Q4_K_M	10.8 / 12.0 GB + 15.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	—
GeForce RTX 2080 Ti	Runs well	est. 44.7 tok/s31.3–58.1theoretical estimate ±30%	Q4_K_M	9.8 / 11.0 GB + 16.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$999 launch MSRP
GeForce RTX 3080 10GB	Runs well	est. 44.6 tok/s31.2–57.9theoretical estimate ±30%	Q4_K_M	8.8 / 10.0 GB + 17.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$699 launch MSRP
GeForce RTX 4070	Runs well	est. 44.4 tok/s31.1–57.7theoretical estimate ±30%	Q4_K_M	10.8 / 12.0 GB + 15.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$599 launch MSRP
GeForce RTX 4070 Super	Runs well	est. 44.4 tok/s31.1–57.7theoretical estimate ±30%	Q4_K_M	10.8 / 12.0 GB + 15.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$599 launch MSRP
Radeon RX 9060 XT 16GB	Runs well	est. 43.9 tok/s30.7–57.0theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$449 street (as of 2026-08-09)
GeForce RTX 4080 Laptop	Runs well	est. 42.6 tok/s29.8–55.3theoretical estimate ±30%	Q4_K_M	10.8 / 12.0 GB + 15.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	—
GeForce GTX 1080 Ti	Runs well	est. 42.3 tok/s29.6–54.9theoretical estimate ±30%	Q4_K_M	9.8 / 11.0 GB + 16.9 GB RAM	MoE experts in RAM	up to 256k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$699 launch MSRP
GeForce RTX 4060 Ti 16GB	Runs well	est. 42.2 tok/s29.5–54.9theoretical estimate ±30%	Q4_K_M	14.8 / 16.0 GB + 11.9 GB RAM	MoE experts in RAM	up to 128k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$499 launch MSRP
GeForce RTX 3060 12GB	Runs well	est. 40.2 tok/s28.2–52.3theoretical estimate ±30%	Q4_K_M	10.8 / 12.0 GB + 15.9 GB RAM	MoE experts in RAM	up to 128k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$250 street (as of 2026-08-09)
Intel Arc B580	Runs well	est. 39.3 tok/s27.5–51.1theoretical estimate ±30%	Q4_K_M	10.8 / 12.0 GB + 15.9 GB RAM	MoE experts in RAM	up to 128k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$290 street (as of 2026-08-09)
GeForce RTX 3070	Runs well	est. 37.3 tok/s26.1–48.5theoretical estimate ±30%	Q4_K_M	6.8 / 8.0 GB + 19.9 GB RAM	MoE experts in RAM	up to 128k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$499 launch MSRP
GeForce RTX 5060	Runs well	est. 37.3 tok/s26.1–48.5theoretical estimate ±30%	Q4_K_M	6.8 / 8.0 GB + 19.9 GB RAM	MoE experts in RAM	up to 128k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$339 street (as of 2026-08-09)
GeForce RTX 5060 Ti 8GB	Runs well	est. 37.3 tok/s26.1–48.5theoretical estimate ±30%	Q4_K_M	6.8 / 8.0 GB + 19.9 GB RAM	MoE experts in RAM	up to 128k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$429 street (as of 2026-08-09)
GeForce RTX 4060 Ti 8GB	Runs well	est. 33.2 tok/s23.2–43.1theoretical estimate ±30%	Q4_K_M	6.8 / 8.0 GB + 19.9 GB RAM	MoE experts in RAM	up to 64k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$399 launch MSRP
GeForce RTX 4060	Runs well	est. 32.6 tok/s22.8–42.3theoretical estimate ±30%	Q4_K_M	6.8 / 8.0 GB + 19.9 GB RAM	MoE experts in RAM	up to 32k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$299 launch MSRP
GeForce RTX 4060 Laptop	Runs well	est. 31.9 tok/s22.3–41.5theoretical estimate ±30%	Q4_K_M	6.8 / 8.0 GB + 19.9 GB RAM	MoE experts in RAM	up to 32k	Not fully on the GPU — Runs great needs the whole model in GPU memory	—
GeForce RTX 4070 Laptop	Runs well	est. 31.9 tok/s22.3–41.5theoretical estimate ±30%	Q4_K_M	6.8 / 8.0 GB + 19.9 GB RAM	MoE experts in RAM	up to 32k	Not fully on the GPU — Runs great needs the whole model in GPU memory	—
GeForce RTX 3050 8GB	Runs well	est. 30.5 tok/s21.3–39.6theoretical estimate ±30%	Q4_K_M	6.8 / 8.0 GB + 19.9 GB RAM	MoE experts in RAM	up to 16k	Not fully on the GPU — Runs great needs the whole model in GPU memory	$249 launch MSRP
Apple M5 Max (32-core GPU) · 36 GB	Runs slowly	est. 80.1 tok/s56.0–104.1theoretical estimate ±30%	Q4_K_M	25.5 GB RAM	CPU only	up to 256k	CPU inference — capped at Runs slowly Try IQ4_XS: Runs great	—
Apple M4 Max (32-core GPU) · 36 GB	Runs slowly	est. 71.4 tok/s49.9–92.8theoretical estimate ±30%	Q4_K_M	25.5 GB RAM	CPU only	up to 256k	CPU inference — capped at Runs slowly Try IQ4_XS: Runs great	—
Apple M6 · 32 GB	Runs slowly	est. 29.7 tok/s20.8–38.6theoretical estimate ±30%	Q4_K_M	25.5 GB RAM	CPU only	up to 256k	CPU inference — capped at Runs slowly Try IQ4_XS: Runs great	—
Apple M5 · 32 GB	Runs slowly	est. 26.7 tok/s18.7–34.8theoretical estimate ±30%	Q4_K_M	25.5 GB RAM	CPU only	up to 256k	CPU inference — capped at Runs slowly Try IQ4_XS: Runs great	—
Apple M4 · 32 GB	Runs slowly	est. 20.9 tok/s14.6–27.1theoretical estimate ±30%	Q4_K_M	25.5 GB RAM	CPU only	up to 256k	CPU inference — capped at Runs slowly Try IQ4_XS: Runs great	—
DDR5-6000 dual-channel CPU · 32 GB	Runs slowly	est. 16.7 tok/s11.7–21.7theoretical estimate ±30%	Q4_K_M	25.5 GB RAM	CPU only	up to 256k	CPU inference — capped at Runs slowly	—
DDR5-5600 dual-channel CPU · 32 GB	Runs slowly	est. 15.6 tok/s10.9–20.3theoretical estimate ±30%	Q4_K_M	25.5 GB RAM	CPU only	up to 256k	CPU inference — capped at Runs slowly	—
DDR4-3200 dual-channel CPU · 32 GB	Runs slowly	est. 8.9 tok/s6.2–11.6theoretical estimate ±30%	Q4_K_M	25.5 GB RAM	CPU only	up to 256k	—	—

Models with a thinking mode spend extra tokens thinking, so their speed thresholds are 1.5× stricter (30 / 12 / 3 tok/s). With thinking turned off, a model may do one grade better.

Cheapest GPUs that run it

Runs greatGeForce RTX 5090$4,200 street (as of 2026-08-09), est. 305.7 tok/s (214.0–397.4, theoretical estimate ±30%).
Runs wellGeForce RTX 3060 12GB$250 street (as of 2026-08-09), est. 40.2 tok/s (28.2–52.3, theoretical estimate ±30%).

Measured results for Nemotron 3.5 Lightning 30B-A3B

No public measurements for this model yet.

Frequently asked questions

How much VRAM does Nemotron 3.5 Lightning 30B-A3B need?

At Q4_K_M the file is 25.48 GB; with 8k context a graphics card needs about 25.1 GB (weights 24.48 GB, KV cache 0.1 GB, compute buffer 0.6 GB). The smallest tracked file (Q2_K (heavy quality loss)) needs about 18.6 GB. Part of the file stays off the graphics card: llama.cpp keeps the input embedding table in system RAM and does not use this model’s MTP (multi-token prediction) block yet, so it never loads it. A Mac counts the whole file, so it needs a little more.

What is the cheapest GPU that runs Nemotron 3.5 Lightning 30B-A3B well?

Cheapest that runs it great: GeForce RTX 5090 at $4,200 (street price as of 2026-08-09), est. 305.7 tok/s (214.0–397.4, theoretical estimate ±30%). Cheaper and still Runs well: GeForce RTX 3060 12GB at $250 (street price as of 2026-08-09), est. 40.2 tok/s (28.2–52.3, theoretical estimate ±30%).

Can Nemotron 3.5 Lightning 30B-A3B run on an 8 GB, 16 GB or 24 GB GPU?

At Q4_K_M with 8k context: GeForce RTX 4060 — Runs well, est. 32.6 tok/s (22.8–42.3, theoretical estimate ±30%); GeForce RTX 5060 Ti 16GB — Runs well, est. 50.8 tok/s (35.6–66.0, theoretical estimate ±30%); GeForce RTX 3090 — Runs well, est. 122.0 tok/s (85.4–158.7, theoretical estimate ±30%).

Speeds are estimates from memory bandwidth, calibrated against public benchmarks, and each one comes with an error band and a confidence label. Real results vary with drivers, backend, context length and thermals.