LLMs ·

Moonlight 16B-A3B Instruct

Parameters
16B (3.0B active)
Architecture
Mixture of experts
Max context
8K tokens
License
mit
Released
2025-02

How to run it

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.

Apps that run Moonlight 16B-A3B Instruct

Our pick for each kind of computer, assuming the model fits in its memory. Pick your own GPU or device in the Model Finder for an exact answer.

Mac (Apple silicon)

Docker Model Runner

Handy if you already work in Docker.

Also good:

Windows PC, NVIDIA GPU

Docker Model Runner

Handy if you already work in Docker.

Also good:

Windows PC, AMD Radeon

Docker Model Runner

Handy if you already work in Docker.

Also good:

Linux, NVIDIA GPU

Docker Model Runner

Handy if you already work in Docker.

Also good:

Linux server, 80 GB+ GPUs

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.

Also good:

App Type Opens it as Systems Engine Start
vLLM Server Original weights WSLWindowsLinux vLLM 
vllm serve moonshotai/Moonlight-16B-A3B-Instruct
SGLang Server Original weights WSLWindowsLinux SGLang 
python3 -m sglang.launch_server --model-path moonshotai/Moonlight-16B-A3B-Instruct
Docker Model Runner CLI + API Original weights macOSWindowsLinux llama.cpp, vLLM 
docker model run hf.co/moonshotai/Moonlight-16B-A3B-Instruct
Text Generation WebUI Desktop app Original weights macOSWindowsLinux llama.cpp, Transformers, ExLlamaV3 Get Text Generation WebUI

Which apps can open a model depends on the files published for it: GGUF for llama.cpp-based apps, MLX for Apple’s engine, the original weights for vLLM and SGLang. Compare every app.

Memory needed

Weights plus KV cache and runtime overhead. Leave a little extra free for the operating system and other apps. Try another context length.

KV cache: latent (MLA) KV cache.

Quantization Quality Weights Total at 8K context Download
Q2_K Low 6.4 GB 7.2 GB All files (opens in a new tab)
Q3_K_M Low 7.8 GB 8.7 GB All files (opens in a new tab)
Q4_K_M Good 9.7 GB 10.6 GB All files (opens in a new tab)
Q5_K_M High 11.3 GB 12.2 GB All files (opens in a new tab)
Q6_K High 13.1 GB 13.9 GB All files (opens in a new tab)
Q8_0 Very high 16.9 GB 17.8 GB All files (opens in a new tab)
F16 Full 31.9 GB 32.8 GB Original weights (opens in a new tab)

Best hardware for Moonlight 16B-A3B Instruct

The least expensive GPU in our catalog that runs it without offloading is the Intel Arc A580 (launch price $179) at Q2_K.

Quality Hardware Quant Memory Speed Runs
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. Sapphire NITRO+ AMD Radeon RX 7900 XTX Vapor-X 24GBAMD Radeon RX 7900 XTX Q8_0 19.0 GB / 24.0 GB 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. ASUS ROG Strix GeForce RTX 3090 OC Edition 24GB GDDR6XNVIDIA GeForce RTX 3090 Q8_0 19.0 GB / 24.0 GB 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 Founders EditionNVIDIA GeForce RTX 4090 Q8_0 19.0 GB / 24.0 GB 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. Sapphire NITRO+ AMD Radeon RX 9070 XT OCAMD Radeon RX 9070 XT Q5_K_M 13.0 GB / 16.0 GB 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 MacBook Pro (14-inch, M4 Pro or M4 Max, 2024)Apple M4 Pro (24 GB) Q5_K_M 13.5 GB / 16.1 GB 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. ASUS Dual GeForce RTX 4060 Ti OC Edition 16GB GDDR6NVIDIA GeForce RTX 4060 Ti 16GB Q5_K_M 13.0 GB / 16.0 GB 72.6 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 4070 Ti SUPERNVIDIA GeForce RTX 4070 Ti SUPER Q5_K_M 13.0 GB / 16.0 GB 127 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 4080 SUPERNVIDIA GeForce RTX 4080 SUPER Q5_K_M 13.0 GB / 16.0 GB 133 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. ASUS Dual GeForce RTX 5060 Ti 16GB GDDR7 OC EditionNVIDIA GeForce RTX 5060 Ti 16GB Q5_K_M 13.0 GB / 16.0 GB 99.0 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. ASUS TUF Gaming GeForce RTX 5070 Ti 16GB GDDR7 OC EditionNVIDIA GeForce RTX 5070 Ti Q5_K_M 13.0 GB / 16.0 GB 147 tok/s Runs great

Check it on your own hardware

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.

Was this page helpful?