whatmodel.sh / EXAONE 4.5 33B
EXAONE 4.5 33B
↗ share
+ contribute
EXAONE 4.5 33B is an open-weights language model from LG AI Research, released April 2026. Its published size is 34.4B parameters and a 262k-token context window. Licence: exaone.
$ how it compares
The LLMs ranked either side of EXAONE 4.5 33B, out of 828 with a score, then more from LG AI Research.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 162 | K-EXAONE (236B-A23B) | 65±1 | local | ||
| 163 |  DeepSeek R1 0528 | 65±1 | api $0.91/M +local | ||
| 164 | EXAONE 4.5 33B | 65±2 | local | ||
| 165 |  Qwen3.5 Flash | 65±1 | api $0.17/M | ||
| 166 |  ChatGPT-4o (2025-03-26) | 65±2 | comparison only | ||
| more from LG AI Research | |||||
| 127 | K-EXAONE 2.0 (750B-A37B) | 67±1 | local | ||
| 251 | EXAONE 4.0 32B | 59±1 | local | ||
| 296 | EXAONE Deep 32B | 56±2 | local | ||
$ what it's good at
agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #116 of 300 * · 21% of its score · 1 test
Beats o3-pro, o3, GPT-5 mini here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Large, Nova 2 Pro (Preview) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| τ²-bench Telecom lab | 73.0% ✎ | ▁·▅█▆▅▃▃▃▂▃▁ beats 57/89 | 21% |
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #139 of 393 * · 5.6% of its score · 1 test
Trails LFM2.5 8B-A1B, LFM2.5 2.6B, Spark-X2.5 1.7B here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| IFEval (aggregation unspecified) lab | 89.6% ✎ | ▄·▄·▄·▄▄···█ beats 5/6 | 5.6% |
math▁▂▃▅▆▅▄█▄▁▁▁ #68 of 398 * · 22% of its score · 1 test
Beats Claude Sonnet 4.6, Claude Opus 4.5, Qwen3.7 Plus here, though they rank higher overall. Trails INTELLECT-3 here, though they rank lower overall.
vision & documentsnot ranked on this task yet · 3 tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| CharXiv (reasoning questions) lab | 71.7% ✎ | ▂▅▃▅▅▆▆▅▆█▆▂ beats 61/100 | — |
| MMMU-Pro standard (SqueezeBits lmms-eval) | 17.6% ✎ | no other model yet | — |
| MMMU-Pro vision (SqueezeBits lmms-eval) | 21.5% ✎ | no other model yet | — |
other tests2 not in a task · 50% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond lab | 80.5% ✎ | █···▄▄···▄·█ beats 6/6 | 34% |
| MMLU-Pro lab +1 | 83.3% ✎ | ▁▁·▁▁▁▃▂▄▅█▁ beats 155/188 | 15% |
$ where to run it
llama.cpp · recommended
version
- 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 | shWindowsirm https://llama.app/install.ps1 | iex - Run EXAONE 4.5 33B · needs ~21.14 GB · from huggingface.co/LGAI-EXAONEchat in the terminal
llama cli -hf LGAI-EXAONE/EXAONE-4.5-33B-GGUF:Q4_K_M - Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf LGAI-EXAONE/EXAONE-4.5-33B-GGUF:Q4_K_M
LM Studio
version
- Install LM Studio · Open the app once before using the lms command.
- Run EXAONE 4.5 33B · needs ~21.14 GB · from huggingface.co/LGAI-EXAONE
- Or download it in the terminal
lms get https://huggingface.co/LGAI-EXAONE/EXAONE-4.5-33B-GGUF@Q4_K_M -y - Then run lms ls to see its name, and lms chat followed by that name
Ollama
version
- Install Ollama
download it (macOS / Windows) ↗
Linuxcurl -fsSL https://ollama.com/install.sh | sh - Run EXAONE 4.5 33B · needs ~21.14 GB · from huggingface.co/LGAI-EXAONEchat in the terminal
ollama run hf.co/LGAI-EXAONE/EXAONE-4.5-33B-GGUF:Q4_K_M
Weights and documentation: huggingface.co/LGAI-EXAONE ↗
$ how the score was worked out · LG AI Research vendor-reporte… 98%
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
| source | results | share of the score |
|---|---|---|
 LG AI Research vendor-reported benchmark scores for its own EXAONE models (Hugging Face model cards) · lab-reported | 5 | 98% |
| prior · pull toward the middle | 2% | |
| weighted average | = 65.0 |
not counted (4)
| SqueezeBits — original EXAONE 4.5 BF16 evaluations | MMLU-Pro [correct] | 72.0% | the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts |
| SqueezeBits — original EXAONE 4.5 BF16 evaluations | MMMU-Pro standard (SqueezeBits lmms-eval) [correct] | 17.6% | fewer than three models have this test yet, or this source ran it for this model alone |
| SqueezeBits — original EXAONE 4.5 BF16 evaluations | MMMU-Pro vision (SqueezeBits lmms-eval) [correct] | 21.5% | fewer than three models have this test yet, or this source ran it for this model alone |
| LG AI Research vendor-reported benchmark scores for its own EXAONE models (Hugging Face model cards) | CharXiv (reasoning questions) [correct] | 71.7% | fewer than three models have this test yet, or this source ran it for this model alone |
$ improve this page
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