FitLocally by GPUwerk

Llama · 3B

Can you run Llama 3.2 3B Instruct locally?

Llama 3.2 3B Instruct is a dense model with 3B parameters. At 4-bit or better it runs on 120 of the 120 hardware configurations listed here; the smallest is the RTX 5060 (8 GB). On a 128 GB DGX Spark it runs at Q4_K_M, at roughly 60 tokens per second (our est.).

Data as of 2 November 2026

Fastest machines for Llama 3.2 3B Instruct

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

  1. 1RTX 5090Graphics card · 32 GB~346 tok/s ModelFit est.Q4_K_M · 3.6 of 31 GB
  2. 2RTX PRO 6000 BlackwellGraphics card · 96 GB~346 tok/s ModelFit est.Q4_K_M · 3.6 of 95 GB
  3. 32x RTX 5090Graphics card · 64 GB~314 tok/s ModelFit est.Q4_K_M · 3.6 of 62 GB
  4. 42x RTX PRO 6000 BlackwellGraphics card · 192 GB~314 tok/s ModelFit est.Q4_K_M · 3.6 of 190 GB
  5. 5RTX 4090Graphics card · 24 GB~239 tok/s ModelFit est.Q4_K_M · 3.6 of 23 GB
  6. 6RTX 5080Graphics card · 16 GB~237 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  7. 7RTX 5070 TiGraphics card · 16 GB~226 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  8. 8M5 UltraMac · 96–512 GB · 5 variants~225 tok/s ModelFit est.Q4_K_M · 3.6 GB
  9. 92x RTX 4090Graphics card · 48 GB~223 tok/s ModelFit est.Q4_K_M · 3.6 of 46 GB
  10. 10RTX 3090 TiGraphics card · 24 GB~219 tok/s ModelFit est.Q4_K_M · 3.6 of 23 GB
  11. 11RTX 3090Graphics card · 24 GB~209 tok/s ModelFit est.Q4_K_M · 3.6 of 23 GB
  12. 122x RTX 3090 TiGraphics card · 48 GB~206 tok/s ModelFit est.Q4_K_M · 3.6 of 46 GB
  13. 132x RTX 3090Graphics card · 48 GB~197 tok/s ModelFit est.Q4_K_M · 3.6 of 46 GB
  14. 14RTX 4080 SUPERGraphics card · 16 GB~191 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  15. 15RTX 4080Graphics card · 16 GB~187 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  16. 16RTX 5070Graphics card · 12 GB~184 tok/s ModelFit est.Q4_K_M · 3.6 of 11.0 GB
  17. 17RTX 4070 Ti SUPERGraphics card · 16 GB~178 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  18. 18RX 7900 XTXGraphics card · 24 GB~166 tok/s ModelFit est.Q4_K_M · 3.6 of 23 GB
  19. 19M5 MaxMac · 48–128 GB · 3 variants~163 tok/s ModelFit est.Q4_K_M · 3.6 GB
  20. 20M3 UltraMac · 96–512 GB · 5 variants~161 tok/s ModelFit est.Q4_K_M · 3.6 GB
  21. 212x RX 7900 XTXGraphics card · 48 GB~159 tok/s ModelFit est.Q4_K_M · 3.6 of 46 GB
  22. 22RX 7900 XTGraphics card · 20 GB~151 tok/s ModelFit est.Q4_K_M · 3.6 of 19.0 GB
  23. 23RTX 4070 TiGraphics card · 12 GB~143 tok/s ModelFit est.Q4_K_M · 3.6 of 11.0 GB
  24. 24RTX 4070 SUPERGraphics card · 12 GB~143 tok/s ModelFit est.Q4_K_M · 3.6 of 11.0 GB
  25. 25RTX 4070Graphics card · 12 GB~143 tok/s ModelFit est.Q4_K_M · 3.6 of 11.0 GB
  26. 26M2 UltraMac · 64–192 GB · 7 variants~139 tok/s ModelFit est.Q4_K_M · 3.6 GB
  27. 27M4 MaxMac · 48–128 GB · 3 variants~137 tok/s ModelFit est.Q4_K_M · 3.6 GB
  28. 28RTX 5060Graphics card · 8 GB~134 tok/s ModelFit est.Q4_K_M · 3.6 of 7.0 GB
  29. 29RTX 5060 TiGraphics card · 16 GB~134 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  30. 30RX 9070 XTGraphics card · 16 GB~132 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  31. 31M5 Max · 36 GB · 32cMac · 36 GB~128 tok/s ModelFit est.Q4_K_M · 3.6 of 24 GB
  32. 32M1 UltraMac · 64–128 GB · 4 variants~124 tok/s ModelFit est.Q4_K_M · 3.6 GB
  33. 33RTX 3070Graphics card · 8 GB~119 tok/s ModelFit est.Q4_K_M · 3.6 of 7.0 GB
  34. 34RTX 3060 TiGraphics card · 8 GB~119 tok/s ModelFit est.Q4_K_M · 3.6 of 7.0 GB
  35. 35M3 MaxMac · 48–128 GB · 3 variants~109 tok/s ModelFit est.Q4_K_M · 3.6 GB
  36. 36RTX 2070Graphics card · 8 GB~108 tok/s ModelFit est.Q4_K_M · 3.6 of 7.0 GB
  37. 37M4 Max · 36 GB · 32cMac · 36 GB~108 tok/s ModelFit est.Q4_K_M · 3.6 of 24 GB
  38. 38M2 MaxMac · 32–96 GB · 5 variants~100 tok/s ModelFit est.Q4_K_M · 3.6 GB
  39. 39RTX 3060Graphics card · 12 GB~99 tok/s ModelFit est.Q4_K_M · 3.6 of 11.0 GB
  40. 40M5 ProMac · 24–64 GB · 6 variants~99 tok/s ModelFit est.Q4_K_M · 3.6 GB
  41. 41M1 MaxMac · 32–64 GB · 4 variants~90 tok/s ModelFit est.Q4_K_M · 3.6 GB
  42. 42RTX 4060 TiGraphics card · 16 GB~87 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  43. 43M3 MaxMac · 36–96 GB · 2 variants~85 tok/s ModelFit est.Q4_K_M · 3.6 GB
  44. 44Tesla P100Graphics card · 16 GB~85 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  45. 45RX 9060 XTGraphics card · 16 GB~84 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  46. 46M4 ProMac · 24–64 GB · 6 variants~83 tok/s ModelFit est.Q4_K_M · 3.6 GB
  47. 47RTX 4060Graphics card · 8 GB~83 tok/s ModelFit est.Q4_K_M · 3.6 of 7.0 GB
  48. 48Tesla T4Graphics card · 16 GB~76 tok/s ModelFit est.Q4_K_M · 3.6 of 15.0 GB
  49. 49Ryzen AI Max+ 395 · 128 GBAI mini PC · 128 GB~63 tok/s ModelFit est.Q4_K_M · 3.6 of 96 GB
  50. 50M2 ProMac · 16–32 GB · 2 variants~61 tok/s ModelFit est.Q4_K_M · 3.6 GB
  51. 51DGX Spark (GB10)AI mini PC · 128 GB~60 tok/s our est.Q4_K_M · 3.6 of 122 GB
  52. 52M6Mac · 24–32 GB · 3 variants~59 tok/s ModelFit est.Q4_K_M · 3.6 GB
  53. 53Ryzen AI Max+ 395 · 64 GBAI mini PC · 64 GB~56 tok/s our est.Q4_K_M · 3.6 of 48 GB
  54. 54M1 ProMac · 16–32 GB · 2 variants~55 tok/s ModelFit est.Q4_K_M · 3.6 GB
  55. 55M6Mac · 16 GB · 2 variants~54 tok/s ModelFit est.Q4_K_M · 3.6 GB
  56. 56M5Mac · 16–32 GB · 6 variants~51 tok/s ModelFit est.Q4_K_M · 3.6 GB
  57. 57M3 ProMac · 18–36 GB · 2 variants~49 tok/s ModelFit est.Q4_K_M · 3.6 GB
  58. 58M4Mac · 16–32 GB · 3 variants~38 tok/s ModelFit est.Q4_K_M · 3.6 GB
  59. 59M3Mac · 8–24 GB · 3 variants~33 tok/s ModelFit est.Q4_K_M · 3.6 GB
  60. 60M2Mac · 8–24 GB · 3 variants~32 tok/s ModelFit est.Q4_K_M · 3.6 GB
  61. 61M1Mac · 8–16 GB · 2 variants~20 tok/s ModelFit est.Q4_K_M · 3.6 GB

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

About the model

Parameters
3B
Active per token
3B
Type
Dense
Good for
Chat

Builds and how to run them

QuantizationLoadMin. RAMCommand
Q4_K_M2.5 GB4 GBollama run llama3.2:3b-instruct-q4_K_M

Load, minimum RAM and Ollama commands: ModelFit, CC BY 4.0. Load is memory with the model loaded, before long contexts.

Questions

Can I run Llama 3.2 3B Instruct 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 M3 · 8 GB.

How much memory does Llama 3.2 3B Instruct need?

About 3.6 GB at the smallest build in our data, including a margin for the runtime and an 8K-token context. Longer contexts need more.

Does Llama 3.2 3B Instruct run on a DGX Spark?

Yes, at Q4_K_M. The Spark has 128 GB of unified memory, about 121.6 GB usable.