Moonlight 16B-A3B Instruct on NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip

Can the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip run Moonlight 16B-A3B Instruct?

Yes, it runs great

Runs great. At F16 it needs about 45.4 GB of 252 GB usable, with an estimated speed of 171 tok/s.

Best quant: F16
Generation speed: 171 tok/s
Prompt processing: 19,200 tok/s
First token (1K prompt): 0.3 s
Largest safe context: 8K

Best app to run it on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip

On Windows

vLLM

Inference server · runs on vLLM

The full model fits in GPU memory at 8-bit, so it can batch many requests: best for serving a team or an app.
On Windows it runs through WSL2.

vllm serve moonshotai/Moonlight-16B-A3B-Instruct

SGLang The full model fits in GPU memory at 8-bit, so it can batch many requests: best for serving a team or an app.
Docker Model Runner Handy if you already work in Docker.

On Linux

vLLM

Inference server · runs on vLLM

The full model fits in GPU memory at 8-bit, so it can batch many requests: best for serving a team or an app.

vllm serve moonshotai/Moonlight-16B-A3B-Instruct

SGLang The full model fits in GPU memory at 8-bit, so it can batch many requests: best for serving a team or an app.
Docker Model Runner Handy if you already work in Docker.

How we choose: the files published for this model, your operating system and GPU, whether it fits fully in memory, and published speed tests (MLX vs llama.cpp on Apple silicon, vLLM vs llama.cpp vs Ollama on NVIDIA). Details.

Download Moonlight 16B-A3B Instruct

Original weights · Hugging Face

Full-precision files from Moonshot AI, for fine-tuning, vLLM, Transformers.

huggingface.co/moonshotai/Moonlight-16B-A3B-Instruct (opens in a new tab)

No sign-in needed.

Links checked 2 days ago against Hugging Face, Ollama and GitHub.

How it’s rated

Quality 46Fair quality (C) · 46 of 100: how capable the model is at F16, whatever it runs on. Match score 64/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at F16, from its editorial quality tier scaled by the quantization or precision. It’s the same on any hardware that runs this variant. Scale: A+ 85+ · A 70+ · B 55+ · C 40+ · D under 40.
Runs Runs great: How well it runs on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip: speed, capped by how well it fits in memory. Runs great: 30+ tokens/s (or near the image, video or decision speed target) · Runs well: 10+ tokens/s · Slow: 3+ tokens/s · Very slow: under 3 · Won’t run: doesn’t fit even with offload. A tight fit or RAM offload caps it at Runs well; heavy offload caps it at Slow.

Match score

To order lists, each result also gets a match score out of 100 from four parts. It favors the most capable model that still runs well.

Model qualityEditorial quality tier of the model, scaled by the quantization or precision used.	20.7 / 45
Memory fitFull marks when everything sits in GPU memory; fewer when part spills to system RAM.	25.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	15.0 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.	4.2 / 15
Match score	64.9 / 100

The match score orders results; it is not shown as a grade. Quality and Runs are the two ratings in the tables.

Every quantization on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	19.8 GB	252 GB	300 tok/s	Runs great	32Basic quality (D) · 32 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 56/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q3_K_M`	21.3 GB	252 GB	285 tok/s	Runs great	38Basic quality (D) · 38 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 59/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q4_K_M`	23.2 GB	252 GB	270 tok/s	Runs great	43Fair quality (C) · 43 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 61/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q5_K_M`	24.8 GB	252 GB	255 tok/s	Runs great	44Fair quality (C) · 44 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 62/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q6_K`	26.5 GB	252 GB	244 tok/s	Runs great	45Fair quality (C) · 45 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 62/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q8_0`	30.4 GB	252 GB	223 tok/s	Runs great	46Fair quality (C) · 46 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`F16`	45.4 GB	252 GB	171 tok/s	Runs great	46Fair quality (C) · 46 of 100: how capable the model is at F16, whatever it runs on. Match score 64/100 (quality, fit, speed and memory use) sets the ranking.	Original weights (opens in a new tab)

Memory figures assume 8K of context. Longer conversations or documents grow the KV cache.

Other hardware that runs it

46Fair quality (C) · 46 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. AMD Radeon RX 7900 XTX Q8_0 · 104 tok/s · Runs great
46Fair quality (C) · 46 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 3090 Q8_0 · 112 tok/s · Runs great
46Fair quality (C) · 46 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4090 Q8_0 · 117 tok/s · Runs great
44Fair quality (C) · 44 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. AMD Radeon RX 9070 XT Q5_K_M · 113 tok/s · Runs great
44Fair quality (C) · 44 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. Apple M4 Pro (24 GB) Q5_K_M · 62.2 tok/s · Runs great
44Fair quality (C) · 44 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4060 Ti 16GB Q5_K_M · 72.6 tok/s · Runs great

Quick answers

How much memory does Moonlight 16B-A3B Instruct need on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip 252GB?

At F16 it needs about 45.4 GB. The NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip 252GB has 252 GB usable.

How fast is Moonlight 16B-A3B Instruct on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip 252GB?

About 171 tok/s at F16, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of Moonlight 16B-A3B Instruct should I use on the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip 252GB?

F16. Of the versions that fit, it is the one that runs best; when two run equally well, the check prefers the higher-rated one.

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.