FitLocally by GPUwerk

NVIDIA · 30B · MoE

Can you run Nemotron 3 Nano 30B-A3B locally?

Nemotron 3 Nano 30B-A3B is a mixture-of-experts model with 30B parameters, 3B active per token. At 4-bit or better it runs on 73 of the 120 hardware configurations listed here; the smallest is the RTX 4090 (24 GB). On a 128 GB DGX Spark it runs at FP16, at roughly 25 tokens per second (our est.).

Data as of 2 November 2026

Fastest machines for Nemotron 3 Nano 30B-A3B

Every machine that runs it, fastest first. Filter by type; your machine is marked. Compare machines running it side by side

  1. 1RTX 5080Graphics card · 16 GB~469 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  2. 2RTX 5070 TiGraphics card · 16 GB~438 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  3. 3RTX 5090Graphics card · 32 GB~398 tok/s our est.Q6_K · 26 of 31 GB
  4. 4RTX 4080 SUPERGraphics card · 16 GB~360 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  5. 5Tesla P100Graphics card · 16 GB~358 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  6. 6RTX 4080Graphics card · 16 GB~351 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  7. 7RTX 4070 Ti SUPERGraphics card · 16 GB~329 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  8. 8RX 9070 XTGraphics card · 16 GB~313 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  9. 92x RTX 5090Graphics card · 64 GB~309 tok/s our est.Q8_0 · 34 of 62 GB
  10. 10RTX 4090Graphics card · 24 GB~305 tok/s our est.Q4_K_M · 19.6 of 23 GB
  11. 11RTX 3090 TiGraphics card · 24 GB~305 tok/s our est.Q4_K_M · 19.6 of 23 GB
  12. 12RX 7900 XTGraphics card · 20 GB~301 tok/s our est.Q3_K_M · 16.0 of 19.0 GB 3-bit only
  13. 13RX 7900 XTXGraphics card · 24 GB~290 tok/s our est.Q4_K_M · 19.6 of 23 GB
  14. 14RTX 3090Graphics card · 24 GB~283 tok/s our est.Q4_K_M · 19.6 of 23 GB
  15. 15RTX 5060 TiGraphics card · 16 GB~219 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  16. 162x RTX 4090Graphics card · 48 GB~174 tok/s our est.Q8_0 · 34 of 46 GB
  17. 172x RTX 3090 TiGraphics card · 48 GB~174 tok/s our est.Q8_0 · 34 of 46 GB
  18. 182x RX 7900 XTXGraphics card · 48 GB~166 tok/s our est.Q8_0 · 34 of 46 GB
  19. 19RTX PRO 6000 BlackwellGraphics card · 96 GB~164 tok/s our est.FP16 · 62 of 95 GB
  20. 202x RTX PRO 6000 BlackwellGraphics card · 192 GB~164 tok/s our est.FP16 · 62 of 190 GB
  21. 212x RTX 3090Graphics card · 48 GB~162 tok/s our est.Q8_0 · 34 of 46 GB
  22. 22Tesla T4Graphics card · 16 GB~156 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  23. 23RX 9060 XTGraphics card · 16 GB~156 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  24. 24RTX 4060 TiGraphics card · 16 GB~141 tok/s our est.Q2_K · 12.5 of 15.0 GB 2-bit only
  25. 25M2 UltraMac · 64 GB · 2 variants~138 tok/s our est.Q8_0 · 34 GB
  26. 26M1 UltraMac · 64 GB · 2 variants~138 tok/s our est.Q8_0 · 34 GB
  27. 27M4 Max · 36 GB · 32cMac · 36 GB~124 tok/s our est.Q4_K_M · 19.6 of 24 GB
  28. 28M2 MaxMac · 32 GB · 2 variants~121 tok/s our est.Q4_K_M · 19.6 GB
  29. 29M1 MaxMac · 32 GB · 2 variants~121 tok/s our est.Q4_K_M · 19.6 GB
  30. 30M4 ProMac · 24 GB · 2 variants~103 tok/s our est.Q3_K_M · 16.0 GB 3-bit only
  31. 31M4 MaxMac · 48–64 GB · 2 variants~94 tok/s our est.Q8_0 · 34 GB
  32. 32M3 Max · 36 GB · 30cMac · 36 GB~91 tok/s our est.Q4_K_M · 19.6 of 24 GB
  33. 33M3 UltraMac · 96–512 GB · 5 variants~75 tok/s our est.FP16 · 62 GB
  34. 34M2 UltraMac · 96–192 GB · 5 variants~73 tok/s our est.FP16 · 62 GB
  35. 35M1 UltraMac · 128 GB · 2 variants~73 tok/s our est.FP16 · 62 GB
  36. 36M3 MaxMac · 48–64 GB · 2 variants~69 tok/s our est.Q8_0 · 34 GB
  37. 37M2 MaxMac · 64 GB · 2 variants~69 tok/s our est.Q8_0 · 34 GB
  38. 38M1 MaxMac · 64 GB · 2 variants~69 tok/s our est.Q8_0 · 34 GB
  39. 39M2 Pro · 32 GBMac · 32 GB~60 tok/s our est.Q4_K_M · 19.6 of 21 GB
  40. 40M1 Pro · 32 GBMac · 32 GB~60 tok/s our est.Q4_K_M · 19.6 of 21 GB
  41. 41M5Mac · 24 GB · 2 variants~58 tok/s our est.Q3_K_M · 16.0 GB 3-bit only
  42. 42M4 Max · 128 GB · 40cMac · 128 GB~50 tok/s our est.FP16 · 62 of 96 GB
  43. 43M4 ProMac · 48–64 GB · 4 variants~47 tok/s our est.Q8_0 · 34 GB
  44. 44M5Mac · 32 GB · 2 variants~46 tok/s our est.Q4_K_M · 19.6 GB
  45. 45M3 Pro · 36 GBMac · 36 GB~45 tok/s our est.Q4_K_M · 19.6 of 24 GB
  46. 46M4 · 24 GBMac · 24 GB~45 tok/s our est.Q3_K_M · 16.0 of 16.0 GB 3-bit only
  47. 47Ryzen AI Max+ 395 · 64 GBAI mini PC · 64 GB~44 tok/s our est.Q8_0 · 34 of 48 GB
  48. 48M3 · 24 GBMac · 24 GB~38 tok/s our est.Q3_K_M · 16.0 of 16.0 GB 3-bit only
  49. 49M2 · 24 GBMac · 24 GB~38 tok/s our est.Q3_K_M · 16.0 of 16.0 GB 3-bit only
  50. 50M2 Max · 96 GB · 38cMac · 96 GB~37 tok/s our est.FP16 · 62 of 72 GB
  51. 51M3 Max · 128 GB · 40cMac · 128 GB~37 tok/s our est.FP16 · 62 of 96 GB
  52. 52M4 · 32 GBMac · 32 GB~36 tok/s our est.Q4_K_M · 19.6 of 21 GB
  53. 53M3 Max · 96 GB · 30cMac · 96 GB~28 tok/s our est.FP16 · 62 of 72 GB
  54. 54DGX Spark (GB10)AI mini PC · 128 GB~25 tok/s our est.FP16 · 62 of 122 GB
  55. 55Ryzen AI Max+ 395 · 128 GBAI mini PC · 128 GB~23 tok/s our est.FP16 · 62 of 96 GB
  56. 56M5 ProMac · 24 GB · 2 variantsno estimateQ3_K_M · 16.0 GB 3-bit only
  57. 57M6Mac · 24 GB · 2 variantsno estimateQ3_K_M · 16.0 GB 3-bit only
  58. 58M6 · 32 GB · 12cMac · 32 GBno estimateQ4_K_M · 19.6 of 21 GB
  59. 59M5 MaxMac · 36–128 GB · 4 variantsno estimateQ4_K_M · 19.6 GB
  60. 60M5 ProMac · 48–64 GB · 4 variantsno estimateQ4_K_M · 19.6 GB
  61. 61M5 UltraMac · 96–512 GB · 5 variantsno estimateQ4_K_M · 19.6 GB
Too big for 24 machines

M3 Pro · 18 GB, RTX 5070, RTX 4070 Ti, RTX 4070 SUPER, RTX 4070, RTX 3060, M2 Pro · 16 GB, M1 Pro · 16 GB, M6 · 16 GB · 12c, M6 · 16 GB · 10c, M5 · 16 GB · 10c, M5 · 16 GB · 8c, M4 · 16 GB, M3 · 16 GB, M2 · 16 GB, M1 · 16 GB, RTX 5060, RTX 4060, RTX 3070, RTX 3060 Ti, RTX 2070, M3 · 8 GB, M2 · 8 GB, M1 · 8 GB. Nemotron 3 Nano 30B-A3B needs about 19.6 GB at its smallest build.

Speeds are single-user decode estimates, not measurements. "ModelFit est." comes from ModelFit's model fitted on published runs (CC BY 4.0); "our est." is memory bandwidth divided by bytes read per token, times 0.55. Memory is what the best build needs, against what the machine can give a model. Method

DGX Spark setup notes: GPUwerk guide for Nemotron 3 Nano 30B-A3B.

About the model

Parameters
30B
Active per token
3B
Type
Mixture of experts
Good for
Fast default with room for context
Model card
Hugging Face

Questions

Can I run Nemotron 3 Nano 30B-A3B on a Mac?

Yes on a Mac with enough memory: the smallest Apple configuration listed here that runs it at 4-bit or better is the M2 Max · 32 GB · 38c.

How much memory does Nemotron 3 Nano 30B-A3B need?

About 12.5 GB at the smallest build in our data, including a margin for the runtime and an 8K-token context. Longer contexts need more.

Does Nemotron 3 Nano 30B-A3B run on a DGX Spark?

Yes, at FP16. The Spark has 128 GB of unified memory, about 121.6 GB usable.