whatmodel.sh / EXAONE 3.5 32B Instruct

EXAONE 3.5 32B Instruct

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

EXAONE 3.5 32B Instruct is an open-weights language model from LG AI Research, released December 2024. Its published size is 32B parameters and a 33k-token context window. Licence: exaone.

Catalogue identity notes

LGAI-EXAONE/EXAONE-3.5-32B-Instruct @3d27aab (created 2024-12-01, EXAONE AI Model License (non-commercial terms)). EXAONE 3.5 bilingual instruct model, largest of the 3.5 family. Lab rows only: rule-(b) rows pooled in first-party-lgai-scores (IFEval and Arena-Hard runs with the catalogued EXAONE models).

Model information sources

Release date ↗ · Weights licence ↗

HF safetensors metadata 32,003,200,000 parameters.

54.9±4.1#321 of 949 LLMsfrom 2 benchmark results across 1 source* thinly tested, treat the rank as provisional

$ how it compares

The LLMs ranked either side of EXAONE 3.5 32B Instruct, out of 1132 with a score, then more from LG AI Research.

#modelscorewhereRAMreleased
422Hunyuan-Standard (2025-02-10) Tencent55±3comparison only2025-02
423PLLuM-12B-nc-chat (2507) PLLuM consortium (CYFRAGOVPL)55±4local~27GB2025-07
424EXAONE 3.5 32B Instruct LG AI Research55±4local~20GB2024-12
425Qwen2.5 32B Instruct Alibaba55±1local~21GB2024-09
426Command A Reasoning (08-2025) Cohere55±2local~34GB
more from LG AI Research
146K-EXAONE 2.0 (750B-A37B) LG AI Research67±1local~791GB2026-07
190K-EXAONE (236B-A23B) LG AI Research65±1local~152GB2025-12
192EXAONE 4.5 33B LG AI Research65±2local~24GB2026-04

see all LLMs · find the ones that fit your machine

$ what it's good at

chat & writing▁▁▁▄▆▅▆█▆▆▆▂ #211 of 466 * · 93% of its score · 2 tests

Beats Granite 4.2 8B, Qwen3-VL 30B-A3B Thinking, Mistral Small 4 (119B-A6.5B) here, though they rank higher overall. Trails LFM2.5 1.2B Instruct, LFM2.5 1.2B Thinking, Gemma 3 12B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena-Hard (version/judge unspecified) lab78.6% ✎▃▃▂▅▃▆▅█▅▅▂▃ beats 26/30easy · avg 68%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 57%small run
IFEval (aggregation unspecified) lab81.7% ✎▄·▄·▄·▄▄···█ beats 3/6easy · avg 78%▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 36%typical

$ where to run it

llama.cpp · recommended

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run EXAONE 3.5 32B Instruct · needs ~19.34 GB · from huggingface.co/LGAI-EXAONE
    chat in the terminal
    llama cli -hf LGAI-EXAONE/EXAONE-3.5-32B-Instruct-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf LGAI-EXAONE/EXAONE-3.5-32B-Instruct-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run EXAONE 3.5 32B Instruct · needs ~19.34 GB · from huggingface.co/LGAI-EXAONE

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/LGAI-EXAONE/EXAONE-3.5-32B-Instruct-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run EXAONE 3.5 32B Instruct · needs ~19 GB · from ollama.com
    chat in the terminal
    ollama run exaone3.5:32b

MLX (Mac)

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run EXAONE 3.5 32B Instruct · needs ~18 GB, Mac only · from huggingface.co/alexander583
    chat in the terminal
    mlx_lm.chat --model alexander583/EXAONE-3.5-32B-Instruct-Q4-mlx
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model alexander583/EXAONE-3.5-32B-Instruct-Q4-mlx

Weights and documentation: huggingface.co/LGAI-EXAONE ↗

$ how the score was worked out · LG AI Research vendor-reporte… 93%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LG AI Research vendor-reported benchmark scores for its own EXAONE models (Hugging Face model cards) · lab-reported2▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 93%Factual vendor-reported score (not licensed content)
prior · pull toward the middle▇▇ 7%
weighted average= 54.9

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json