[whatmodel.sh](../../README.md) / EXAONE 4.0 1.2B # EXAONE 4.0 1.2B LG AI Research · released 2025-07 · 1.28B params · exaone **score 47.0** ±1.4 · #436 of 732 · from 9 benchmark results across 1 source · † no independently run results verified Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at ● marks EXAONE 4.0 1.2B among every model on that task or test, worst to best. 
task            vs the others          rank  counts  tests
agent tasks     ▁▁▃●█▄▅▅▄▄▂▁    #282 of 313     15%      3
chat & writing  ▁▁▁▃▆●▆█▆▇▇▃   #291 of 402*    6.2%      1
math            ▁▁▃▅●▆▄█▄▂▁▁   #281 of 431*     35%      2
agent tasks · #282 of 313 · 15% of its score Beats [Magistral Small 1.0 (2506)](magistral-small-2506.md), [Llama 4 Scout (109B-A17B)](llama-4-scout.md), [Bielik 11B v2.3 Instruct](bielik-11b-v2-3-instruct.md) here, though they rank higher overall. Trails [Granite 4.0 H 350M](granite-4-0-h-350m.md), [Qwen3 0.6B](qwen3-0-6b.md), [Qwen2.5 1.5B Instruct](qwen2-5-1-5b.md) here, though they rank lower overall. 
test                     result  vs the others in its run   difficulty        counts  why
τ-bench Retail (maker)    28.1%  ▁····●·▂▇█▂▆ beats 1/28    medium · avg 38%    7.7%  at its level · small run
τ-bench Airline (maker)   20.5%  ●▁▁·▅▄▁█▇▁▁▇ beats 0/28    hard · avg 28%      5.0%  small run
BFCL v3 (maker)           52.9%  ▁▁▁▁▁▃▅●▆▅▄▁ beats 58/171  medium · avg 57%    2.4%  blunt · at its level · small run
chat & writing · #291 of 402 · 6.2% of its score Beats [Ministral 3 14B Reasoning 2512](ministral-3-14b-reasoning.md), [Ministral 3 8B](ministral-3-8b.md), [Gemma 4 E4B](gemma-4-e4b.md) here, though they rank higher overall. Trails [EXAONE 3.5 2.4B Instruct](exaone-3-5-2-4b.md), [LFM2 700M](lfm2-700m.md), [Flan-T5 XXL](flan-t5-xxl.md) here, though they rank lower overall. 
test                                result  vs the others in its run  difficulty      counts  why
IFEval (aggregation unspecified) …   67.8%  ●·▄·▄·▄▄···█ beats 0/6    easy · avg 78%    6.2%  at its level
math · #281 of 431 · 35% of its score Beats [Molmo 2 8B](molmo-2-8b.md), [Llama 3.3 70B Instruct](llama-3-3-70b.md), [DeepSeek-V2.5](deepseek-v2-5.md) here, though they rank higher overall. Trails [DeepSeek R1 Distill Qwen 1.5B](deepseek-r1-distill-qwen-1-5b.md), [Gemma 7B (base)](gemma-7b-base.md), [Tulu 2 DPO 13B](tulu-2-dpo-13b.md) here, though they rank lower overall. 
test                        result  vs the others in its run  difficulty        counts  why
AIME 2025 (maker)            45.2%  ●·▄▄····▄··█ beats 0/6    medium · avg 54%     27%  sharp · at its level
HMMT February 2025 (maker)   34.0%  ▁▁●▃▁▅▅▃██▁▄ beats 2/31   medium · avg 51%    7.5%  at its level · small run
other tests (4) 
test                             result  vs the others in its run   difficulty        counts  why
GPQA Diamond (maker)              52.0%  ●···▄▄···▄·█ beats 0/6     hard · avg 51%       23%  sharp · at its level
MMLU-Pro (maker)                  59.3%  ▁▁·▁▁▁▃●▄▅█▁ beats 60/188  medium · avg 62%     14%  at its level
MMLU-Redux (maker)                71.5%  ▁▁●▁▂▃▄▅▇▅█▄ beats 3/63    easy · avg 79%      6.0%  at its level · small run
Thunder-KoNUBench symbol 0-shot   26.5%  ●··········█ beats 0/1                            —  not counted yet: no other model in the …
## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's 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 is added up by who ran the tests. [Method](../method.md) 
source                                    results  share of the score
LG AI Research vendor-reported benchmar…        9   99% ▇▇▇▇▇▇▇▇▇▇▇▇
prior: pull toward the middle                        1% ▇
weighted average                                   = 47.0
Not counted (1) 
source                  test                            score  why
Thunder-KoNUBench: A …  Thunder-KoNUBench symbol 0-sh…  26.5%  too few models have it yet
## How to run it ### llama.cpp (recommended) 1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md) ```sh # macOS / Linux curl -LsSf https://llama.app/install.sh | sh # Windows irm https://llama.app/install.ps1 | iex ``` 2. Run EXAONE 4.0 1.2B (gguf-q4_k_m, needs ~1.27 GB, from [huggingface.co/LGAI-EXAONE](https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF/tree/main)) ```sh llama cli -hf LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF:Q4_K_M ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF:Q4_K_M ``` Other versions: gguf-q8_0 (~2.04 GB) `llama cli -hf LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF:Q8_0` ### LM Studio 1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download) 2. Run EXAONE 4.0 1.2B (gguf-q4_k_m, needs ~1.27 GB, from [huggingface.co/LGAI-EXAONE](https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF/tree/main)) ```sh lms get https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF@Q4_K_M -y ``` 3. Then run lms ls to see its name, and lms chat followed by that name Other versions: gguf-q8_0 (~2.04 GB) `lms get https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF@Q8_0 -y` · 4bit (~0.7 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/EXAONE-4.0-1.2B-MLX-4bit -y` · 8bit (~1.4 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/EXAONE-4.0-1.2B-MLX-8bit -y` ### Ollama 1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download) ```sh # Linux curl -fsSL https://ollama.com/install.sh | sh ``` 2. Run EXAONE 4.0 1.2B (gguf-q4_k_m, needs ~1.27 GB, from [huggingface.co/LGAI-EXAONE](https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF/tree/main)) ```sh ollama run hf.co/LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF:Q4_K_M ``` Other versions: gguf-q8_0 (~2.04 GB) `ollama run hf.co/LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF:Q8_0` ### MLX (Mac) 1. Install MLX (Mac). Apple silicon Macs only, text only. ```sh # with uv uv tool install mlx-lm # or pip install mlx-lm ``` 2. Run EXAONE 4.0 1.2B (4bit, needs ~0.7 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/EXAONE-4.0-1.2B-MLX-4bit/tree/da095e728aab942c3ade54860b76f510be4054f4)) ```sh mlx_lm.chat --model lmstudio-community/EXAONE-4.0-1.2B-MLX-4bit ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model lmstudio-community/EXAONE-4.0-1.2B-MLX-4bit ``` Other versions: 8bit (~1.4 GB, Mac only) `mlx_lm.chat --model lmstudio-community/EXAONE-4.0-1.2B-MLX-8bit` Model repository and documentation: [huggingface.co/LGAI-EXAONE](https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-1.2B) [interactive version](https://whatmodel.sh/model/exaone-4-0-1-2b) · [JSON](https://whatmodel.sh/api?format=json&all=1)