Outmake Models
Models

K2-Horizon 36B

Institute of Foundation Models · September 2026

GetHugging Face

For problems worked through step by step, with its thinking shown apart from the final answer.

Rating, estimated
4.2
Downloads
38K
Parameters
37B, 4B active
Context
512K
Download
11 to 75 GB
Commercial use
Yes

Tight fit on M4 · 16 GB

Outmake picks the IQ2_XXS size: a 11 GB download that uses 12 GB of memory, and writes about 6.2 words (8.3 tokens) a second.

About

The sparse member of the K2-Horizon family, so only part of the model runs for each word it writes.

Good to know

Its training data, training recipe and training code are all public, along with the model.

Sources: IFM/K2-Horizon-MoVA-36B-A4B model card, Outmake's speed estimates, LM Arena and Hugging Face. Every line comes from one of these.

Good at

  • Tasks that call outside tools
  • Reading long documents

Licence

Apache 2.0. Commercial use: yes. The maker's page has the full terms.

K2-Horizon 36B runs well from a Mac with an M3 Pro and 18 GB of memory, at IQ2_XXS, writing about 7.7 words (10 tokens) a second. 48 of 59 Mac configurations run it well.

Compare with

The nearest models in size, judged for the same Mac.

Sizes

SizeDownloadMemory at 8KMemory at 32KOpen
Q3_K_MSmaller17 GB18 GB21 GBOpen
Q4_K_MBalanced23 GB24 GB26 GBOpen
Q6_KBest32 GB33 GB36 GBOpen
All 24 sizes
SizeDownloadMemory at 8KMemory at 32KOpen
IQ2_XXS11 GB12 GBestimated14 GBOpen
IQ2_XS12 GB13 GBestimated15 GBOpen
IQ2_S12 GB13 GBestimated16 GBOpen
IQ2_M13 GB14 GBestimated17 GBOpen
Q2_K14 GB15 GBestimated17 GBOpen
IQ3_XXS16 GB17 GBestimated20 GBOpen
Q3_K_S17 GB18 GBestimated20 GBOpen
IQ3_XS17 GB18 GBestimated21 GBOpen
Q3_K_L18 GB19 GBestimated22 GBOpen
IQ3_M18 GB19 GBestimated22 GBOpen
IQ4_XS20 GB21 GBestimated24 GBOpen
IQ4_NL22 GB23 GBestimated25 GBOpen
Q4_022 GB23 GBestimated25 GBOpen
Q4_K_S22 GB23 GBestimated26 GBOpen
Q4_K_L23 GB24 GBestimated27 GBOpen
Q4_124 GB25 GBestimated27 GBOpen
Q5_K_S26 GB27 GBestimated30 GBOpen
Q5_K_M27 GB28 GBestimated30 GBOpen
Q6_K_L32 GB33 GBestimated36 GBOpen
Q8_040 GB41 GBestimated43 GBOpen
BF1675 GB76 GBestimated78 GBOpen

Memory is what llama.cpp's own calculator reported for each size when the catalog was built, with the app's settings. Smaller, Balanced and Best are the three sizes the store names.

Run It Anywhere

K2-Horizon 36B is an open model, so any app that runs GGUF files can run it. These take the Q4_K_M size from bartowski straight from Hugging Face.

  • LM StudioMac, Windows, LinuxOpen
  • JanMac, Windows, LinuxOpen
  • MstyMac, Windows, LinuxOpen
  • OllamaMac, Windows, Linuxollama run hf.co/bartowski/K2-Horizon-MoVA-36B-A4B-GGUF:Q4_K_M
  • llama.cppMac, Windows, Linuxllama-server -hf bartowski/K2-Horizon-MoVA-36B-A4B-GGUF:Q4_K_M
  • Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/bartowski/K2-Horizon-MoVA-36B-A4B-GGUF:Q4_K_M
On a PC Graphics Card
Video memoryLargest size that fitsDownloadMemory at 8K
6 GBNone worth runningneeds 17 GB
8 GBNone worth runningneeds 17 GB
10 GBNone worth runningneeds 17 GB
11 GBNone worth runningneeds 17 GB
12 GBNone worth runningneeds 17 GB
16 GBNone worth runningneeds 17 GB
20 GBIQ3_M18 GB19 GB
24 GBQ4_K_S22 GB23 GB
32 GBQ5_K_M27 GB28 GB
48 GBQ8_040 GB41 GB

The largest size, of those that keep three bits a weight or more, whose memory at an 8K chat is under 92% of the card's; the rest is the desktop's. Estimated from the files, the same way as for a Mac; speed on a PC is not estimated yet. A size that doesn't fit the card still runs, slower, with part of it in ordinary memory.

On every Mac
MacSizeFitMemoryWritesReads
M6 · 16 GBIQ2_XXSTight fit12 GB9.21100
M6 · 24 GBQ2_KRuns well15 GB8.81100
M6 · 32 GBQ3_K_MRuns well21 GB8.41100
M5 · 16 GBIQ2_XXSTight fit12 GB8.5406
M5 · 24 GBQ2_KRuns well15 GB8.1406
M5 · 32 GBQ3_K_MRuns well21 GB7.7406
M5 Pro · 24 GBQ2_KRuns well15 GB16813
M5 Pro · 48 GBQ3_K_MRuns great21 GB15813
M5 Pro · 64 GBQ4_K_MRuns great26 GB141016
M5 Max · 36 GBQ4_K_MRuns well24 GB201625
M5 Max · 48 GBQ3_K_MRuns great21 GB282032
M5 Max · 64 GBQ4_K_MRuns great26 GB262032
M5 Max · 128 GBQ6_KRuns great60 GB242032
M5 Ultra · 96 GBQ6_KRuns great46 GB443250
M5 Ultra · 256 GBQ6_KRuns great60 GB444063
M5 Ultra · 512 GBQ6_KRuns great60 GB444063
M4 · 16 GBIQ2_XXSTight fit12 GB6.2137
M4 · 24 GBQ2_KRuns well15 GB6137
M4 · 32 GBQ3_K_MRuns well21 GB5.7137
M4 Pro · 24 GBQ2_KRuns well15 GB12273
M4 Pro · 48 GBQ3_K_MRuns great21 GB12273
M4 Pro · 64 GBQ4_K_MRuns great26 GB11273
M4 Max · 36 GBQ4_K_MRuns well24 GB15546
M4 Max · 48 GBQ3_K_MRuns great21 GB20683
M4 Max · 64 GBQ4_K_MRuns great26 GB19683
M4 Max · 96 GBQ6_KRuns great46 GB14546
M4 Max · 128 GBQ6_KRuns great60 GB17683
M3 · 8 GBQ3_K_MToo big46 GB
M3 · 16 GBIQ2_XXSTight fit12 GB5.4113
M3 · 24 GBQ2_KRuns well15 GB5.1113
M3 Pro · 18 GBIQ2_XXSRuns well12 GB7.7198
M3 Pro · 36 GBQ4_K_MRuns well24 GB6.7198
M3 Max · 36 GBQ4_K_MRuns well24 GB12424
M3 Max · 48 GBQ3_K_MRuns great21 GB16565
M3 Max · 64 GBQ4_K_MRuns great26 GB15565
M3 Max · 96 GBQ6_KRuns great46 GB11424
M3 Max · 128 GBQ6_KRuns great60 GB14565
M3 Ultra · 96 GBQ6_KRuns great46 GB24848
M3 Ultra · 256 GBQ6_KRuns great60 GB24848
M3 Ultra · 512 GBQ6_KRuns great60 GB241130
M2 · 8 GBQ3_K_MToo big46 GB
M2 · 16 GBIQ2_XXSTight fit12 GB5.3111
M2 · 24 GBQ2_KRuns well15 GB5.1111
M2 Pro · 16 GBIQ2_XXSTight fit12 GB9.9222
M2 Pro · 32 GBQ3_K_MRuns well21 GB9.2222
M2 Max · 32 GBQ3_K_MRuns well21 GB16416
M2 Max · 64 GBQ4_K_MRuns great26 GB15416
M2 Max · 96 GBQ6_KRuns great46 GB14527
M2 Ultra · 64 GBQ4_K_MRuns great26 GB25832
M2 Ultra · 128 GBQ6_KRuns great60 GB24832
M2 Ultra · 192 GBQ6_KRuns great60 GB24832
M1 · 8 GBQ3_K_MToo big46 GB
M1 · 16 GBIQ2_XXSTight fit12 GB3.690
M1 Pro · 16 GBIQ2_XXSTight fit12 GB9.1180
M1 Pro · 32 GBQ3_K_MRuns well21 GB8.5180
M1 Max · 32 GBQ3_K_MRuns well21 GB14308
M1 Max · 64 GBQ4_K_MRuns great26 GB13308
M1 Ultra · 64 GBQ4_K_MRuns great26 GB21617
M1 Ultra · 128 GBQ6_KRuns great60 GB19617

Writes and reads are words a second (a token is about three quarters of a word), 4K tokens into a chat, estimated unless marked measured. The size is the one the store offers that Mac. Constants fitted October 11, 2026; how, and how wrong.

Badge for the model card

Shows on K2-Horizon 36B's Hugging Face card which Macs run it well, linked to this page. It updates every morning.

Runs on Mac badge

Sizes and files are read from bartowski/K2-Horizon-MoVA-36B-A4B-GGUF on Hugging Face and checked every morning. LM Arena has not rated this model, so the rating is Outmake's estimate from its size and release date; how the estimate is made. Catalog of November 4, 2026.

Your Computer

Every model is judged for this Mac. Both are under About This Mac, in the Apple menu.

Every Mac · Every graphics card

Civitai

Open in OutmakeHugging Face

New to Outmake? Download the free beta