FitLocally by GPUwerk

Kolibri · 78.1B

Can you run Kolibri 1 locally?

Kolibri 1 is a 78.1B-parameter model. At 4-bit or better it runs on 47 of the 120 hardware configurations listed here; the smallest is the 2x RTX 4090 (48 GB). On a 128 GB DGX Spark it runs at IQ4_XS.

Data as of 2 November 2026

Fastest machines for Kolibri 1

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

  1. 1RTX PRO 6000 BlackwellGraphics card · 96 GB~282 tok/s ModelFit est.Q4_K_M · 52 of 95 GB
  2. 22x RTX 5090Graphics card · 64 GB~260 tok/s ModelFit est.Q4_K_M · 52 of 62 GB
  3. 32x RTX PRO 6000 BlackwellGraphics card · 192 GB~260 tok/s ModelFit est.Q4_K_M · 52 of 190 GB
  4. 4M5 UltraMac · 96–512 GB · 5 variants~181 tok/s ModelFit est.Q4_K_M · 52 GB
  5. 5M5 Max · 64 GB · 40cMac · 64 GB~135 tok/s ModelFit est.IQ4_XS · 46 of 48 GB
  6. 6M3 UltraMac · 96–512 GB · 5 variants~128 tok/s ModelFit est.Q4_K_M · 52 GB
  7. 7M2 UltraMac · 64 GB · 2 variants~121 tok/s ModelFit est.IQ4_XS · 46 GB
  8. 8M5 Max · 128 GB · 40cMac · 128 GB~121 tok/s ModelFit est.Q4_K_M · 52 of 96 GB
  9. 9M4 Max · 64 GB · 40cMac · 64 GB~113 tok/s ModelFit est.IQ4_XS · 46 of 48 GB
  10. 10M2 UltraMac · 96–192 GB · 5 variants~112 tok/s ModelFit est.Q4_K_M · 52 GB
  11. 11M1 UltraMac · 64 GB · 2 variants~108 tok/s ModelFit est.IQ4_XS · 46 GB
  12. 12M4 Max · 128 GB · 40cMac · 128 GB~102 tok/s ModelFit est.Q4_K_M · 52 of 96 GB
  13. 13M1 UltraMac · 128 GB · 2 variants~100 tok/s ModelFit est.Q4_K_M · 52 GB
  14. 14M3 Max · 64 GB · 40cMac · 64 GB~90 tok/s ModelFit est.IQ4_XS · 46 of 48 GB
  15. 15M2 MaxMac · 64 GB · 2 variants~83 tok/s ModelFit est.IQ4_XS · 46 GB
  16. 16M5 ProMac · 64 GB · 2 variants~80 tok/s ModelFit est.IQ4_XS · 46 GB
  17. 17M3 Max · 128 GB · 40cMac · 128 GB~80 tok/s ModelFit est.Q4_K_M · 52 of 96 GB
  18. 18M1 MaxMac · 64 GB · 2 variants~75 tok/s ModelFit est.IQ4_XS · 46 GB
  19. 19M2 Max · 96 GB · 38cMac · 96 GB~75 tok/s ModelFit est.Q4_K_M · 52 of 72 GB
  20. 20M4 ProMac · 64 GB · 2 variants~67 tok/s ModelFit est.IQ4_XS · 46 GB
  21. 21M3 Max · 96 GB · 30cMac · 96 GB~62 tok/s ModelFit est.Q4_K_M · 52 of 72 GB
  22. 22Ryzen AI Max+ 395 · 128 GBAI mini PC · 128 GB~47 tok/s ModelFit est.Q4_K_M · 52 of 96 GB
  23. 232x RTX 4090Graphics card · 48 GBno estimateIQ4_XS · 46 of 46 GB
  24. 242x RTX 3090 TiGraphics card · 48 GBno estimateIQ4_XS · 46 of 46 GB
  25. 252x RTX 3090Graphics card · 48 GBno estimateIQ4_XS · 46 of 46 GB
  26. 262x RX 7900 XTXGraphics card · 48 GBno estimateIQ4_XS · 46 of 46 GB
  27. 27Ryzen AI Max+ 395 · 64 GBAI mini PC · 64 GBno estimateIQ4_XS · 46 of 48 GB
  28. 28DGX Spark (GB10)AI mini PC · 128 GBno estimateIQ4_XS · 46 of 122 GB
Too big for 73 machines

M5 Max · 48 GB · 40c, M4 Max · 48 GB · 40c, M3 Max · 48 GB · 40c, M5 Pro · 48 GB · 20c, M5 Pro · 48 GB · 16c, M4 Pro · 48 GB · 20c, M4 Pro · 48 GB · 16c, RTX 5090, M5 Max · 36 GB · 32c, M4 Max · 36 GB · 32c, M3 Max · 36 GB · 30c, M3 Pro · 36 GB, RTX 4090, RTX 3090 Ti, RTX 3090, RX 7900 XTX, M2 Max · 32 GB · 38c, M2 Max · 32 GB · 30c, M1 Max · 32 GB · 32c, M1 Max · 32 GB · 24c, M2 Pro · 32 GB, M1 Pro · 32 GB, M6 · 32 GB · 12c, M5 · 32 GB · 10c, M5 · 32 GB · 8c, M4 · 32 GB, RX 7900 XT, M5 Pro · 24 GB · 20c, M5 Pro · 24 GB · 16c, M4 Pro · 24 GB · 20c, M4 Pro · 24 GB · 16c, M6 · 24 GB · 12c, M6 · 24 GB · 10c, M5 · 24 GB · 10c, M5 · 24 GB · 8c, M4 · 24 GB, M3 · 24 GB, M2 · 24 GB, RTX 5080, RTX 5070 Ti, RTX 5060 Ti, RTX 4080 SUPER, RTX 4080, RTX 4070 Ti SUPER, RTX 4060 Ti, Tesla T4, Tesla P100, RX 9070 XT, RX 9060 XT, 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. Kolibri 1 needs about 46 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

About the model

Parameters
78.1B
Active per token
not in our sources
Type
not in our sources
Good for
Reasoning, German, Long context

Builds and how to run them

QuantizationLoadMin. RAMCommand
Q4_K_M50 GB96 GBGGUF or MLX build
IQ4_XS44 GB64 GBGGUF or MLX build

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 Kolibri 1 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 Ultra · 64 GB · 76c.

How much memory does Kolibri 1 need?

About 46 GB at the smallest build in our data, including a margin for the runtime and an 8K-token context. Longer contexts need more.

Does Kolibri 1 run on a DGX Spark?

Yes, at IQ4_XS. The Spark has 128 GB of unified memory, about 121.6 GB usable.