Outmake Models
Models

Internlm2 Math Plus 20B

Shanghai AI Lab · May 2024

Get Hugging Face

A 20B chat model from Shanghai AI Lab.

Downloads
8.2K
Likes
2
Parameters
20B
Context
8K
Licence
other

Runs well on M4 · 16 GB

The IQ3_M size from legraphista is the one to get: a 9.1 GB download that uses 10 GB of memory, and writes about 4.4 words a second.

Ratings

KnowledgeMMLU-Pro
37.1%

MMLU-Pro: graduate-level questions in 14 subjects. Scores are the ones filed on the model's Hugging Face page or the board's own table, read 2026-11-05; some come from the maker.

Internlm2 Math Plus 20B runs well from a Mac with an M1 and 16 GB of memory. 56 of 59 Mac configurations run it well.

Compare with

The nearest models in size.

Sizes

From legraphista/internlm2-math-plus-20b-IMat-GGUF on Hugging Face. A smaller size is a smaller download and a little less accurate.

SizeDownloadMemory at 8KGet
IQ1_S4.5 GB5.6 GBestimatedOpen
IQ1_M4.9 GB6.0 GBestimatedOpen
IQ2_XXS5.5 GB6.6 GBestimatedOpen
IQ2_XS6.1 GB7.1 GBestimatedOpen
IQ2_S6.5 GB7.5 GBestimatedOpen
IQ2_M7.0 GB8.0 GBestimatedOpen
Q2_K_S7.0 GB8.1 GBestimatedOpen
Q2_K7.5 GB8.6 GBestimatedOpen
IQ3_XXS7.8 GB8.9 GBestimatedOpen
IQ3_XS8.4 GB9.4 GBestimatedOpen
Q3_K_S8.8 GB9.8 GBestimatedOpen
IQ3_S8.8 GB9.8 GBestimatedOpen
IQ3_M9.1 GB10 GBestimatedOpen
Q3_K9.7 GB11 GBestimatedOpen
Q3_K_L11 GB12 GBestimatedOpen
IQ4_XS11 GB12 GBestimatedOpen
IQ4_NL11 GB12 GBestimatedOpen
Q4_K_S11 GB12 GBestimatedOpen
Q4_K12 GB13 GBestimatedOpen
Q5_K_S14 GB15 GBestimatedOpen
Q5_K14 GB15 GBestimatedOpen
Q6_K16 GB17 GBestimatedOpen
Q8_021 GB22 GBestimatedOpen
BF1640 GB41 GBestimatedOpen
internlm2-math-plus-20b.FP1640 GB41 GBestimatedOpen

Run It Anywhere

Internlm2 Math Plus 20B is an open model, so any app that runs GGUF files can run it. These take the IQ3_M size from legraphista straight from Hugging Face.

  • LM StudioMac, Windows, LinuxOpen
  • JanMac, Windows, LinuxOpen
  • MstyMac, Windows, LinuxOpen
  • OllamaMac, Windows, Linuxollama run hf.co/legraphista/internlm2-math-plus-20b-IMat-GGUF:IQ3_M
  • llama.cppMac, Windows, Linuxllama-server -hf legraphista/internlm2-math-plus-20b-IMat-GGUF:IQ3_M
  • Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/legraphista/internlm2-math-plus-20b-IMat-GGUF:IQ3_M
On a PC Graphics Card
Video memoryLargest size that fitsDownloadMemory at 8K
6 GBNone worth runningneeds 8.9 GB
8 GBNone worth runningneeds 8.9 GB
10 GBIQ3_S8.8 GB9.8 GB
11 GBQ3_K9.7 GB11 GB
12 GBIQ4_XS11 GB12 GB
16 GBQ5_K14 GB15 GB
20 GBQ6_K16 GB17 GB
24 GBQ8_021 GB22 GB
32 GBQ8_021 GB22 GB
48 GBinternlm2-math-plus-20b.FP1640 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
MacSizeFitMemoryWrites
M6 · 16 GBIQ3_MRuns well10 GB6.2
M6 · 24 GBQ3_K_LRuns great12 GB5.7
M6 · 32 GBQ5_KRuns great15 GB4.7
M5 · 16 GBIQ3_MRuns well10 GB5.6
M5 · 24 GBQ3_K_LRuns great12 GB5.2
M5 · 32 GBQ5_KRuns great15 GB4.3
M5 Pro · 24 GBQ3_K_LRuns great12 GB10
M5 Pro · 48 GBQ8_0Runs great22 GB7.5
M5 Pro · 64 GBQ8_0Runs great22 GB7.5
M5 Max · 36 GBQ6_KRuns great17 GB12
M5 Max · 48 GBQ8_0Runs great22 GB15
M5 Max · 64 GBQ8_0Runs great22 GB15
M5 Max · 128 GBQ8_0Runs great22 GB15
M5 Ultra · 96 GBQ8_0Runs great22 GB28
M5 Ultra · 256 GBQ8_0Runs great22 GB28
M5 Ultra · 512 GBQ8_0Runs great22 GB28
M4 · 16 GBIQ3_MRuns well10 GB4.4
M4 · 24 GBQ3_K_LRuns great12 GB4
M4 · 32 GBQ5_KRuns great15 GB3.4
M4 Pro · 24 GBQ3_K_LRuns great12 GB8.7
M4 Pro · 48 GBQ8_0Runs great22 GB6.3
M4 Pro · 64 GBQ8_0Runs great22 GB6.3
M4 Max · 36 GBQ6_KRuns great17 GB9.6
M4 Max · 48 GBQ8_0Runs great22 GB12
M4 Max · 64 GBQ8_0Runs great22 GB12
M4 Max · 96 GBQ8_0Runs great22 GB9.2
M4 Max · 128 GBQ8_0Runs great22 GB12
M3 · 8 GBIQ1_STight fit5.6 GB5.2
M3 · 16 GBIQ3_MRuns well10 GB3.7
M3 · 24 GBQ3_K_LRuns great12 GB3.4
M3 Pro · 18 GBIQ4_XSRuns well12 GB5
M3 Pro · 36 GBQ6_KRuns great17 GB3.8
M3 Max · 36 GBQ6_KRuns great17 GB7.3
M3 Max · 48 GBQ8_0Runs great22 GB9.1
M3 Max · 64 GBQ8_0Runs great22 GB9.1
M3 Max · 96 GBQ8_0Runs great22 GB7
M3 Max · 128 GBQ8_0Runs great22 GB9.1
M3 Ultra · 96 GBQ8_0Runs great22 GB17
M3 Ultra · 256 GBQ8_0Runs great22 GB17
M3 Ultra · 512 GBQ8_0Runs great22 GB17
M2 · 8 GBIQ1_STight fit5.6 GB5.2
M2 · 16 GBIQ3_MRuns well10 GB3.7
M2 · 24 GBQ3_K_LRuns great12 GB3.4
M2 Pro · 16 GBIQ3_MRuns well10 GB7.2
M2 Pro · 32 GBQ5_KRuns great15 GB5.5
M2 Max · 32 GBQ5_KRuns great15 GB10
M2 Max · 64 GBQ8_0Runs great22 GB9.1
M2 Max · 96 GBQ8_0Runs great22 GB9.1
M2 Ultra · 64 GBQ8_0Runs great22 GB17
M2 Ultra · 128 GBQ8_0Runs great22 GB17
M2 Ultra · 192 GBQ8_0Runs great22 GB17
M1 · 8 GBIQ1_STight fit5.6 GB3.5
M1 · 16 GBIQ3_MRuns well10 GB2.5
M1 Pro · 16 GBIQ3_MRuns well10 GB6.8
M1 Pro · 32 GBQ5_KRuns great15 GB5.3
M1 Max · 32 GBQ5_KRuns great15 GB9.8
M1 Max · 64 GBQ8_0Runs great22 GB8.7
M1 Ultra · 64 GBQ8_0Runs great22 GB15
M1 Ultra · 128 GBQ8_0Runs great22 GB15

Writes is words a second, estimated from the Mac's memory speed and the model's size; how, and how wrong.

Packaged by

2 repositories on Hugging Face carry this model for llama.cpp. Outmake reads them all as one model.

Information

Maker
Shanghai AI Lab
Released
May 2024
Parameters
20B
Longest chat
8K tokens
Reads pictures
No
Licence
other
Downloads this month
8,238
Likes
2
Sizes
25, 4.5 to 40 GB
Kind
Maker's release

Badge for the model card

Shows on Internlm2 Math Plus 20B's Hugging Face card which Macs run it well, linked to this page. It updates every morning.

Runs on Mac badge

Files, downloads and likes read from Hugging Face on November 5, 2026 and refreshed every morning. Memory and speed are estimates until someone running Outmake measures this model on that Mac.

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