RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

Can I run Moonlight 16B A3B Instruct locally?

Moonlight 16B A3B Instruct by moonshotai has 16B parameters (3.0B active per token, 64 experts, 6 used per token). At 4-bit with an 8K context it needs about 10.8 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track.

10.8 GBQ4_K_M · 8K context
RTX 4060 Ti 16GBsmallest consumer GPU that fits
M2 24GBsmallest Mac that fits
8Kmax context
MLAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from mmnga/Moonlight-16B-A3B-Instruct-gguf where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx
2-bit (Q2_K) •rough7.0 GB7.1 GB
3-bit (Q3_K_M) •usable8.5 GB8.7 GB
4-bit (Q4_K_M) •good, most popular10.6 GB10.8 GB
5-bit (Q5_K_M) •very good12.0 GB12.2 GB
6-bit (Q6_K) •near lossless14.1 GB14.2 GB
8-bit (Q8_0) •effectively lossless16.6 GB16.8 GB
16-bit (F16/BF16)original weights30.6 GB30.7 GB

Which GPUs and Macs run Moonlight 16B A3B Instruct?

Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBFitsQ3_K_M8.7 GB231details
RTX 4060 Ti 16GB16 GBFitsQ5_K_M12.2 GB127details
RTX 309024 GBFitsQ8_016.8 GB293details
RTX 409024 GBFitsQ8_016.8 GB316details
RTX 509032 GBFitsQ8_016.8 GB562details
RTX 5070 Ti16 GBFitsQ5_K_M12.2 GB396details
RX 7900 XTX24 GBFitsQ8_016.8 GB301details
Mac M4 Pro 48GB48 GBFitsQ8_016.8 GB86details
Mac M4 Max 128GB128 GBFitsF1630.7 GB91details
Mac M3 Ultra 512GB512 GBFitsF1630.7 GB137details
DGX Spark128 GBFitsF1630.7 GB46details
RTX 407012 GBFitsQ3_K_M8.7 GB324details
RTX 5060 Ti 16GB16 GBFitsQ5_K_M12.2 GB198details
RTX 40608 GBTightQ2_K7.1 GB220details
Arc B58012 GBFitsQ3_K_M8.7 GB293details
RTX 3060 Ti8 GBTightQ2_K7.1 GB362details
RTX 30708 GBTightQ2_K7.1 GB362details
RTX 3080 10GB10 GBFitsQ2_K7.1 GB615details
RTX 3080 12GB12 GBFitsQ3_K_M8.7 GB586details
RTX 3090 Ti24 GBFitsQ8_016.8 GB316details
RTX 4060 Ti 8GB8 GBTightQ2_K7.1 GB233details
RTX 4070 Super12 GBFitsQ3_K_M8.7 GB324details
RTX 4070 Ti12 GBFitsQ3_K_M8.7 GB324details
RTX 4070 Ti Super16 GBFitsQ5_K_M12.2 GB297details
RTX 408016 GBFitsQ5_K_M12.2 GB317details
RTX 4080 Super16 GBFitsQ5_K_M12.2 GB325details
RTX 50608 GBTightQ2_K7.1 GB362details
RTX 507012 GBFitsQ3_K_M8.7 GB431details
RTX 508016 GBFitsQ5_K_M12.2 GB424details
2x RTX 30902×24 GBFitsF1630.7 GB156details
2x RTX 40902×24 GBFitsF1630.7 GB168details
2x RTX 50902×32 GBFitsF1630.7 GB299details
RTX A600048 GBFitsF1630.7 GB128details
RTX 6000 Ada48 GBFitsF1630.7 GB160details
RTX PRO 6000 Blackwell96 GBFitsF1630.7 GB299details
L40S48 GBFitsF1630.7 GB144details
A100 40GB40 GBFitsF1630.7 GB259details
A100 80GB80 GBFitsF1630.7 GB340details
H100 PCIe80 GBFitsF1630.7 GB333details
H100 SXM80 GBFitsF1630.7 GB558details
H200141 GBFitsF1630.7 GB800details
RX 76008 GBTightQ2_K7.1 GB233details
RX 7700 XT12 GBFitsQ3_K_M8.7 GB277details
RX 7800 XT16 GBFitsQ5_K_M12.2 GB276details
RX 7900 GRE16 GBFitsQ5_K_M12.2 GB255details
RX 7900 XT20 GBFitsQ8_016.8 GB251details
RX 907016 GBFitsQ5_K_M12.2 GB283details
RX 9070 XT16 GBFitsQ5_K_M12.2 GB283details
Ryzen AI Max+ 395 128GB128 GBFitsF1630.7 GB43details
Arc A770 16GB16 GBFitsQ5_K_M12.2 GB247details
Mac M1 16GB16 GBFitsQ3_K_M8.7 GB44details
Mac M2 24GB24 GBFitsQ6_K14.2 GB37details
Mac M3 24GB24 GBFitsQ6_K14.2 GB37details
Mac M4 16GB16 GBFitsQ3_K_M8.7 GB77details
Mac M4 32GB32 GBFitsQ8_016.8 GB38details
Mac M1 Pro 32GB32 GBFitsQ8_016.8 GB63details
Mac M3 Pro 36GB36 GBFitsQ8_016.8 GB47details
Mac M4 Pro 24GB24 GBFitsQ6_K14.2 GB102details
Mac M4 Pro 64GB64 GBFitsF1630.7 GB46details
Mac M1 Max 64GB64 GBFitsF1630.7 GB67details
Mac M2 Max 96GB96 GBFitsF1630.7 GB67details
Mac M3 Max 128GB128 GBFitsF1630.7 GB67details
Mac M4 Max 64GB64 GBFitsF1630.7 GB91details
Mac M2 Ultra 192GB192 GBFitsF1630.7 GB133details
Mac M3 Ultra 256GB256 GBFitsF1630.7 GB137details

Questions

How much VRAM does Moonlight 16B A3B Instruct need?

About 10.8 GB at 4-bit (Q4_K_M) with an 8K context: 9.8 GB of weights (real GGUF file size), 0.2 GB of KV cache and 0.7 GB of runtime overhead. At 8-bit it needs about 16.8 GB.

What is the smallest GPU that runs Moonlight 16B A3B Instruct?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 4060 Ti 16GB (16 GB).

Can I run Moonlight 16B A3B Instruct on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M2 24GB or larger, using the default macOS GPU memory limit.

Model facts

Hugging Face repomoonshotai/Moonlight-16B-A3B-Instruct
Parameters15,960,111,936 (16B) · ~3.0B active (from the model name)
Layers / hidden size27 / 2048
AttentionMLA; 16 heads, 16 KV heads, head dim 128
KV cache per token (FP16)30.4 KiB
Max context8,192 tokens
Licensemit
Released on HFFeb 22, 2025
Downloads (30d) / likes34,522 / 211

Sources: Hugging Face model API · config.json (moonshotai/Moonlight-16B-A3B-Instruct) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.