readygpu

Nemotron 3 Super 120B-A12B

123.6B total, 12B active · 256k context · nvidia-nemotron-open-model-license · released 2026-03-10

chatcodingagenticreasoning

What to run it on in the US

Fully in memory at 128k context, priced at retailers today.

Something we price holds it fully in memory, but none reaches 15 tok/s, our bar for recommending it. The table below has the speeds.

All hardware that runs it

Everything we price, best value first. Two cards split the model by layers.

HardwareQuantFitsDecodePricetok/s per 1,000
Mac Studio M4 Max 128GBUD-Q4_K_Mfits23 tok/s––
Framework Desktop Ryzen AI Max+ 395 128GBUD-Q4_K_Mfits9 tok/s––
Mac Studio M3 Ultra 512GBQ8_0fits23 tok/s––
NVIDIA DGX Spark 128GBUD-Q6_Kfits8 tok/s––
GMKtec EVO-X2 Ryzen AI Max+ 395 128GBUD-Q4_K_Mfits9 tok/s––
Mac Studio M3 Ultra 256GBQ8_0fits23 tok/s––
Mac Studio M3 Ultra 96GBUD-Q3_K_Mfits44 tok/s––
NVIDIA RTX PRO 6000 BlackwellUD-Q4_K_Mfits72 tok/s––
ASUS Ascent GX10 128GBUD-Q6_Kfits8 tok/s––
NVIDIA GeForce RTX 5090UD-Q2_K_XLoffload11 tok/s––
AMD Radeon AI PRO R9700UD-Q2_K_XLoffload10 tok/s––

29 other cards and machines we price cannot hold it at this context.

Memory needed

Weights, KV cache and runtime buffers at each context length (KV cache at fp16).

QuantFile@8k@32k@64k@128k
UD-Q4_K_Mgguf76.9 GB77.8 GB78.2 GB78.6 GB79.5 GB
UD-Q2_K_XLgguf50.9 GB51.9 GB52.2 GB52.6 GB53.5 GB
UD-Q3_K_Mgguf57.5 GB58.4 GB58.8 GB59.2 GB60.1 GB
UD-Q5_K_Mgguf100.0 GB100.9 GB101.3 GB101.7 GB102.6 GB
UD-Q6_Kgguf106.9 GB107.8 GB108.2 GB108.6 GB109.5 GB
Q8_0gguf119.6 GB120.6 GB120.9 GB121.4 GB122.3 GB

Run it

Copy a command for the runtime you use.

  • UD-Q4_K_M
    • Ollamaollama run hf.co/unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q4_K_M
    • LM Studiolms get unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF@ud-q4_k_m
    • llama.cppllama-server -hf unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q4_K_M
  • UD-Q2_K_XL
    • Ollamaollama run hf.co/unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q2_K_XL
    • LM Studiolms get unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF@ud-q2_k_xl
    • llama.cppllama-server -hf unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q2_K_XL
  • UD-Q3_K_M
    • Ollamaollama run hf.co/unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q3_K_M
    • LM Studiolms get unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF@ud-q3_k_m
    • llama.cppllama-server -hf unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q3_K_M
  • UD-Q5_K_M
    • Ollamaollama run hf.co/unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q5_K_M
    • LM Studiolms get unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF@ud-q5_k_m
    • llama.cppllama-server -hf unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q5_K_M
  • UD-Q6_K
    • Ollamaollama run hf.co/unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q6_K
    • LM Studiolms get unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF@ud-q6_k
    • llama.cppllama-server -hf unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-Q6_K
  • Q8_0
    • Ollamaollama run hf.co/unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:Q8_0
    • LM Studiolms get unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF@q8_0
    • llama.cppllama-server -hf unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:Q8_0