FitLocally by GPUwerk

Gemma · 2.3B

Can you run Gemma 4 E2B locally?

Gemma 4 E2B is a 2.3B-parameter model. 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.

Data as of 2 November 2026

Fastest machines for Gemma 4 E2B

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~380 tok/s ModelFit est.Q8_0 · 5.7 of 31 GB
  2. 2RTX PRO 6000 BlackwellGraphics card · 96 GB~380 tok/s ModelFit est.Q8_0 · 5.7 of 95 GB
  3. 32x RTX 5090Graphics card · 64 GB~341 tok/s ModelFit est.Q8_0 · 5.7 of 62 GB
  4. 42x RTX PRO 6000 BlackwellGraphics card · 192 GB~341 tok/s ModelFit est.Q8_0 · 5.7 of 190 GB
  5. 5RTX 4090Graphics card · 24 GB~269 tok/s ModelFit est.Q8_0 · 5.7 of 23 GB
  6. 6RTX 5080Graphics card · 16 GB~268 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  7. 7RTX 5070 TiGraphics card · 16 GB~256 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  8. 82x RTX 4090Graphics card · 48 GB~249 tok/s ModelFit est.Q8_0 · 5.7 of 46 GB
  9. 9M5 UltraMac · 96–512 GB · 5 variants~248 tok/s ModelFit est.Q8_0 · 5.7 GB
  10. 10RTX 3090 TiGraphics card · 24 GB~246 tok/s ModelFit est.Q8_0 · 5.7 of 23 GB
  11. 11RTX 3090Graphics card · 24 GB~235 tok/s ModelFit est.Q8_0 · 5.7 of 23 GB
  12. 122x RTX 3090 TiGraphics card · 48 GB~229 tok/s ModelFit est.Q8_0 · 5.7 of 46 GB
  13. 132x RTX 3090Graphics card · 48 GB~220 tok/s ModelFit est.Q8_0 · 5.7 of 46 GB
  14. 14RTX 4080 SUPERGraphics card · 16 GB~218 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  15. 15RTX 4080Graphics card · 16 GB~214 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  16. 16RTX 5070Graphics card · 12 GB~210 tok/s ModelFit est.Q8_0 · 5.7 of 11.0 GB
  17. 17RTX 4070 Ti SUPERGraphics card · 16 GB~204 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  18. 18M5 MaxMac · 48–128 GB · 3 variants~188 tok/s ModelFit est.Q8_0 · 5.7 GB
  19. 19RX 7900 XTXGraphics card · 24 GB~182 tok/s ModelFit est.Q8_0 · 5.7 of 23 GB
  20. 20M3 UltraMac · 96–512 GB · 5 variants~180 tok/s ModelFit est.Q8_0 · 5.7 GB
  21. 212x RX 7900 XTXGraphics card · 48 GB~173 tok/s ModelFit est.Q8_0 · 5.7 of 46 GB
  22. 22RX 7900 XTGraphics card · 20 GB~166 tok/s ModelFit est.Q8_0 · 5.7 of 19.0 GB
  23. 23RTX 4070 TiGraphics card · 12 GB~165 tok/s ModelFit est.Q8_0 · 5.7 of 11.0 GB
  24. 24RTX 4070 SUPERGraphics card · 12 GB~165 tok/s ModelFit est.Q8_0 · 5.7 of 11.0 GB
  25. 25RTX 4070Graphics card · 12 GB~165 tok/s ModelFit est.Q8_0 · 5.7 of 11.0 GB
  26. 26M4 MaxMac · 48–128 GB · 3 variants~158 tok/s ModelFit est.Q8_0 · 5.7 GB
  27. 27RTX 5060Graphics card · 8 GB~155 tok/s ModelFit est.Q8_0 · 5.7 of 7.0 GB
  28. 28RTX 5060 TiGraphics card · 16 GB~155 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  29. 29M2 UltraMac · 64–192 GB · 7 variants~154 tok/s ModelFit est.Q8_0 · 5.7 GB
  30. 30M5 Max · 36 GB · 32cMac · 36 GB~150 tok/s ModelFit est.Q8_0 · 5.7 of 24 GB
  31. 31RX 9070 XTGraphics card · 16 GB~146 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  32. 32RTX 3070Graphics card · 8 GB~138 tok/s ModelFit est.Q8_0 · 5.7 of 7.0 GB
  33. 33RTX 3060 TiGraphics card · 8 GB~138 tok/s ModelFit est.Q8_0 · 5.7 of 7.0 GB
  34. 34M1 UltraMac · 64–128 GB · 4 variants~136 tok/s ModelFit est.Q8_0 · 5.7 GB
  35. 35RTX 2070Graphics card · 8 GB~127 tok/s ModelFit est.Q8_0 · 5.7 of 7.0 GB
  36. 36M3 MaxMac · 48–128 GB · 3 variants~127 tok/s ModelFit est.Q8_0 · 5.7 GB
  37. 37M4 Max · 36 GB · 32cMac · 36 GB~126 tok/s ModelFit est.Q8_0 · 5.7 of 24 GB
  38. 38M5 ProMac · 24–64 GB · 6 variants~116 tok/s ModelFit est.Q8_0 · 5.7 GB
  39. 39M2 MaxMac · 32–96 GB · 5 variants~116 tok/s ModelFit est.Q8_0 · 5.7 GB
  40. 40RTX 3060Graphics card · 12 GB~116 tok/s ModelFit est.Q8_0 · 5.7 of 11.0 GB
  41. 41M1 MaxMac · 32–64 GB · 4 variants~104 tok/s ModelFit est.Q8_0 · 5.7 GB
  42. 42RTX 4060 TiGraphics card · 16 GB~102 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  43. 43M3 MaxMac · 36–96 GB · 2 variants~100 tok/s ModelFit est.Q8_0 · 5.7 GB
  44. 44Tesla P100Graphics card · 16 GB~99 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  45. 45M4 ProMac · 24–64 GB · 6 variants~98 tok/s ModelFit est.Q8_0 · 5.7 GB
  46. 46RTX 4060Graphics card · 8 GB~97 tok/s ModelFit est.Q8_0 · 5.7 of 7.0 GB
  47. 47RX 9060 XTGraphics card · 16 GB~96 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  48. 48Tesla T4Graphics card · 16 GB~90 tok/s ModelFit est.Q8_0 · 5.7 of 15.0 GB
  49. 49Ryzen AI Max+ 395 · 128 GBAI mini PC · 128 GB~73 tok/s ModelFit est.Q8_0 · 5.7 of 96 GB
  50. 50M2 ProMac · 16–32 GB · 2 variants~72 tok/s ModelFit est.Q8_0 · 5.7 GB
  51. 51M6Mac · 24–32 GB · 3 variants~70 tok/s ModelFit est.Q8_0 · 5.7 GB
  52. 52M1 ProMac · 16–32 GB · 2 variants~64 tok/s ModelFit est.Q8_0 · 5.7 GB
  53. 53M6Mac · 16 GB · 2 variants~64 tok/s ModelFit est.Q8_0 · 5.7 GB
  54. 54M5Mac · 16–32 GB · 6 variants~61 tok/s ModelFit est.Q8_0 · 5.7 GB
  55. 55M3 ProMac · 18–36 GB · 2 variants~59 tok/s ModelFit est.Q8_0 · 5.7 GB
  56. 56M3 · 8 GBMac · 8 GB~55 tok/s ModelFit est.Q4_K_M · 3.4 of 5.3 GB
  57. 57M2 · 8 GBMac · 8 GB~51 tok/s ModelFit est.Q4_K_M · 3.4 of 5.3 GB
  58. 58M4Mac · 16–32 GB · 3 variants~45 tok/s ModelFit est.Q8_0 · 5.7 GB
  59. 59M3Mac · 16–24 GB · 2 variants~40 tok/s ModelFit est.Q8_0 · 5.7 GB
  60. 60M2Mac · 16–24 GB · 2 variants~37 tok/s ModelFit est.Q8_0 · 5.7 GB
  61. 61M1 · 8 GBMac · 8 GB~33 tok/s ModelFit est.Q4_K_M · 3.4 of 5.3 GB
  62. 62M1 · 16 GBMac · 16 GB~24 tok/s ModelFit est.Q8_0 · 5.7 of 10.7 GB
  63. 63Ryzen AI Max+ 395 · 64 GBAI mini PC · 64 GBno estimateQ4_K_M · 3.4 of 48 GB
  64. 64DGX Spark (GB10)AI mini PC · 128 GBno estimateQ4_K_M · 3.4 of 122 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
2.3B
Active per token
not in our sources
Type
not in our sources
Good for
IoT, Mobile, Edge

Builds and how to run them

QuantizationLoadMin. RAMCommand
Q4_K_M2.3 GB4 GBollama run gemma4:e2b
Q8_04.6 GB8 GBollama run gemma4:e2b-it-q8_0

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 Gemma 4 E2B 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 Gemma 4 E2B need?

About 3.4 GB at the smallest build in our data, including a margin for the runtime and an 8K-token context. Longer contexts need more.

Does Gemma 4 E2B run on a DGX Spark?

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