llmrun
OsaurusAINemotronNemotronHForCausalLM

Nemotron 3.5 Lightning 30B A3B JANG 4M — Hardware Requirements & GPU Compatibility

Chat

Nemotron 3.5 Lightning 30B A3B JANG 4M is a 32.9B-parameter open language model from OsaurusAI in the Nemotron family. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 20.12 GB of VRAM — see which GPUs and Macs can run it below.

32.9B params · MoE3.7B active262K contextOther295 downloads3 likes
Based on NVIDIA Nemotron 3.5 Lightning 30B A3B BF16
S

Runs great on a 24 GB GPU

Runs from16 GB Q2_K · 256KBest on24 GB Q4_K_M · ~159 tok/s · 256KMacs from18 GB Q2_K · ~39 tok/s

Other

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run Nemotron 3.5 Lightning 30B A3B JANG 4M

One recommended way. Every other method, OS and quant is in the step-by-step guide.

Recommended

vLLM
one command · needs vLLM installed
$vllm serve OsaurusAI/Nemotron-3.5-Lightning-30B-A3B-JANG_4M --max-model-len 16384

vLLM repo with its own quantisation

Download: Weights on Hugging Face

install, OS, other quants

Other ways

each opens step-by-step

Run in cloud

Fits on RTX A6000 (48 GB) (27 GB headroom) · Q4_K_M

Generation speed
~135 tok/s
generation speed
Cost per 1M output tokens
$0.68
per 1M output tokens
Deployon RunPodBest optionfrom $0.33/hr (opens in new tab)
Compare GPUs →
or

How Much VRAM Does Nemotron 3.5 Lightning 30B A3B JANG 4M Need?

Select a quantization level; the listed hardware adheres to it.

QuantizationVRAM+32KOn RTX 4090 (reference)
14.4 GB14.4 GB2-bit · K-quant — 2-bit quantization with K-quant improvementsFits: 256K context · ~188 tok/s
16.4 GB16.5 GB3-bit medium — 3-bit medium quantizationFits: 256K context · ~176 tok/s
20.1 GB20.2 GB4-bit medium — 4-bit medium quantization — most popular sweet spotFits: 256K context · ~159 tok/s
23.8 GB23.9 GB5-bit medium — 5-bit medium quantization — good quality/size tradeoffFits: 32K context · ~144 tok/s
27.5 GB27.6 GB6-bit, very good — 6-bit quantization, very good qualityDoes not fit: needs 27.5 GB
33.3 GB33.4 GB8-bit, near-lossless — 8-bit quantization, near-losslessDoes not fit: needs 33.3 GB
66.2 GB66.3 GB16-bit · bfloat — Brain floating point 16 — preferred for trainingDoes not fit: needs 66.2 GB
est.calculated, no published file yet

Selected Q4_K_M

4-bit medium quantization

20.1GBat 16K context

Weights
19.8 GB
KV cache
0.1 GB
Overhead
0.3 GB

Fits RTX 4090 (reference) · 3.9 GB free · ~159 tok/s

≥50 tok/s needs ≥203 GB/s · RTX 5090 ~208, RTX 3090 Ti ~159

Which Hardware Can Run Nemotron 3.5 Lightning 30B A3B JANG 4M?

Q4_K_M · 20.1 GB

Nemotron 3.5 Lightning 30B A3B JANG 4M (Q4_K_M) requires 20.1 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 27+ GB is recommended. Using the full 262K context window can add up to 1.2 GB, bringing total usage to 21.3 GB. 7 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 122 devices with unified memory can run Nemotron 3.5 Lightning 30B A3B JANG 4M, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M1 Max (32 GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M
8–12 GB · RTX 4060, RTX 3060 12GBDoesn't fit — smallest quant needs 14.4 GB8 GB tier →12 GB tier →
16 GB · RTX 5060 Ti 16GBQ2_K · 14.4 GB · ~119 tok/s · 256K16 GB tier →
24 GB · RTX 4090referenceQ4_K_M · 20.1 GB · ~159 tok/s · 256K24 GB tier →
Mac 32 GB · Mac Mini M4Q4_K_M · 20.1 GB · ~27 tok/s · 256KMacs →
Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 20.1 GB · ~61 tok/s · 256KMacs →

Alternatives to Nemotron 3.5 Lightning 30B A3B JANG 4M

Nemotron family · 42 models →

Similar models

6
ModelScoreVRAMtok/sWhy it is similar
MiniCPM V 4.7 35B A3BOpenBMB—22.1 GB~169same size
Llm Jp 4.1 32B A3b Thinkingllm-jp—20.1 GB~188same size
Llm Jp 4.1 33B Thinkingllm-jp—22.5 GB~35same size
Agnes 3.0 FlashAgnes-AI—22.1 GB~35same size
Agents A1InternScience—21.5 GB~171same size
Ornith 1.0 35BDeep Reinforce—21.5 GB~171same size

Specifications

Publisher
OsaurusAI
Family
Nemotron
Parameters
32.9B
Active parameters
3.7B per token
Architecture
NemotronHForCausalLM
Context Length
262,144 tokens
Vocabulary Size
131,072
Release Date
2026-08-11
License
Other

Resources

Frequently Asked Questions

How much VRAM does Nemotron 3.5 Lightning 30B A3B JANG 4M need?

Nemotron 3.5 Lightning 30B A3B JANG 4M requires 20.1 GB of VRAM at Q4_K_M, or 66.2 GB at BF16. Full 262K context adds up to 1.2 GB (21.3 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 32.9B × 4.8 bits ÷ 8 = 19.7 GB

KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 1.6 GB (at full 262K context)

VRAM usage by quantization

20.1 GB
21.3 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run Nemotron 3.5 Lightning 30B A3B JANG 4M?

Yes, at Q5_K_M (23.8 GB) or lower. Higher quantizations like Q6_K (27.5 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for Nemotron 3.5 Lightning 30B A3B JANG 4M?

For Nemotron 3.5 Lightning 30B A3B JANG 4M, Q4_K_M (20.1 GB) offers the best balance of quality and VRAM usage. Q5_K_M (23.8 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 14.4 GB.

VRAM requirement by quantization

Q2_K
14.4 GB
Q4_K_M ★
20.1 GB
Q5_K_M
23.8 GB
Q6_K
27.5 GB
Q8_0
33.3 GB
BF16
66.2 GB

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Nemotron 3.5 Lightning 30B A3B JANG 4M on a Mac?

Yes — MacBook Pro 14" M3 Pro (18 GB) and 108 other Macs can run Nemotron 3.5 Lightning 30B A3B JANG 4M, the smallest at Q2_K. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 20.1 GB of usable unified memory (RAM minus macOS overhead), from Mac Mini M5 Pro (24 GB).

Can I run Nemotron 3.5 Lightning 30B A3B JANG 4M locally?

Yes — Nemotron 3.5 Lightning 30B A3B JANG 4M can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Nemotron 3.5 Lightning 30B A3B JANG 4M?

At Q4_K_M, Nemotron 3.5 Lightning 30B A3B JANG 4M can reach ~92 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~159 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)

Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.

Example: NVIDIA B200 → 2.3 GB active ÷ (8000 × 0.65) = 0.43 ms, plus 52 layers × 0.055 ms = 2.86 ms, so 1000 ÷ 3.29 ms = ~304 tok/s

Estimated speed at Q4_K_M (20.1 GB)

~304 tok/s
~159 tok/s
~304 tok/s
~280 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of Nemotron 3.5 Lightning 30B A3B JANG 4M?

At Q4_K_M, the download is about 19.75 GB. The full-precision BF16 version is 65.83 GB. The smallest option (Q2_K) is 13.99 GB.

Which GPUs can run Nemotron 3.5 Lightning 30B A3B JANG 4M?

7 consumer GPUs can run Nemotron 3.5 Lightning 30B A3B JANG 4M at Q4_K_M. Top options include NVIDIA GeForce RTX 5090, AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090. 1 GPU has plenty of headroom for comfortable inference.

Which devices can run Nemotron 3.5 Lightning 30B A3B JANG 4M?

122 devices with unified memory can run Nemotron 3.5 Lightning 30B A3B JANG 4M at Q4_K_M (20.1 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.