KORMo-Web-4B

KORMo-Web-4B is a 4B vision-language agent that completes real browser tasks from screenshots alone. Give it a request in Korean or English; it plans, clicks, types and moves between sites and tabs until the task is done. In the same live-web evaluation, it beats every open-weight web agent we compared, by +9.6 points in English and +22.1 points in Korean.

On Online-Mind2Web it reaches 42.3% in English and 42.3% in Korean, which is +22.3 / +28.8 points over the base Qwen3.5-4B.

What Makes KORMo-Web Different?

  • 10M+ Atomic-Level Web Pretraining Samples. Large-scale Korean and English web understanding pretraining teaches the model what individual UI elements are, where they are, and what they do—covering text, location, appearance, layout, function, and state.
  • 100K+ Web Interaction Trajectories. Bilingual browser demonstrations teach the model to translate fine-grained web understanding into real-world, multi-step actions.
  • Open Models and Data. Releasing model weights, large-scale bilingual web pretraining data, interaction trajectories, and evaluation benchmarks to support reproducible web agent research.

Performance

All models below ran the same 612 live-website tasks (300 English, 312 Korean), each with its own prompt and action format, and were scored by the same LLM judge.

Model Size Online-Mind2Web EN Online-Mind2Web KO
Qwen3.5-4B (base, no fine-tuning) 4B 20.0 13.5
Fara-1.5-4B 4B 20.7 7.7
UI-TARS-1.5-7B 7B 22.3 20.2
OpenWebRL-4B 4B 32.7 15.4
KORMo-Web-4B (ours) 4B 42.3 42.3
KORMo-Web-9B (ours) 9B 51.0 48.1

Success rate (%). The denominator is always the full task count, so crashed or empty runs count as failures.

KORMo-Web-4B outperforms open web agents of similar size in both English and Korean.

Why KORMo-Web

  • Korean and English, natural requests. Handles short commands as well as long, conversational instructions in either language.
  • Screenshot-only. It needs no DOM or accessibility tree; it acts on what is on screen, like a person, and checks each action's result on the next screenshot.
  • Thinks, then acts. A short reasoning step precedes every action, so its behaviour is easy to inspect and debug.
  • Multi-site, multi-tab workflows. It can search on one site, verify on another and write the result into a third, managing tabs on its own.
  • Open and ready to run. Apache-2.0 weights, a reference browser harness, the tool schema and the exact inference settings are all included.

How to Start: Chrome Extension

The fastest way to try KORMo-Web is the KORMo-Web Chrome extension: type a request in Chrome's side panel and watch the agent work in your own browser. This video walks through the whole setup, from serving the model to running a task and controlling it from Discord.

Setup guide (about 15 min, Korean narration). Parts that showed account tokens have been blurred.

1. Serve the model on a Mac or on an NVIDIA GPU.

Mac (Apple silicon)

KORMo-Web-4B (4-bit or full precision) and KORMo-Web-9B all run on a Mac with vllm-mlx 0.6.0 or later. Pick the one that fits your Mac's memory:

Model Weights Memory in use Recommended Mac memory
KORMo-Web-4B-mlx-4bit (4-bit) 2.8 GB about 3.8 GB 16 GB or more
KORMo-Web-4B (bf16) 9.6 GB about 9.5 GB 24 GB or more
KORMo-Web-9B (bf16) 17.5 GB about 19 GB 48 GB or more (32 GB is tight)

Measured with vllm-mlx 0.6.0 after a few short requests (macOS units, 1 GB = 1024³ bytes). Long episodes add KV cache, about 1 GB per 30k tokens, and Chrome needs memory too. macOS lets the GPU use only about 2/3–3/4 of RAM.

uv venv --python 3.12 && source .venv/bin/activate
uv pip install "vllm-mlx>=0.6.0" huggingface_hub
export HF_HUB_DISABLE_XET=1    # only if large downloads stall

MODEL=KORMo-Web-4B             # or KORMo-Web-4B-mlx-4bit / KORMo-Web-9B
hf download KORMo-VL/$MODEL --local-dir ~/models/$MODEL
vllm-mlx serve ~/models/$MODEL --port 8010 --served-model-name KORMo-VL/$MODEL \
  --enable-auto-tool-choice --tool-call-parser auto

Keep this terminal open. --served-model-name is the name clients must request; without it the server registers the local folder path and requests by model name return 404. vllm-mlx 0.6.0 and later detect the vision model by itself; older versions need --mllm.

NVIDIA GPU (local PC or server)

Serve the model with vLLM. Much faster, and one server can be shared.

pip install vllm
vllm serve KORMo-VL/KORMo-Web-4B --host 0.0.0.0 --port 8000 \
  --served-model-name KORMo-Web-4B \
  --mm-encoder-tp-mode data --mm-processor-cache-type shm \
  --gdn-prefill-backend triton

Check that it is up with curl http://localhost:8010/v1/models (Mac) or curl http://<server>:8000/v1/models (vLLM). If you changed the port, use that port.

2. Install the extension. git clone https://github.com/MLP-Lab/KORMo-Web-Extension.git, open chrome://extensions, turn on Developer mode, click Load unpacked and choose the cloned folder. Open the GUI 에이전트 side panel, type a task and press 실행 (Run).

3. Point the extension at your model. Open the panel's 개발자 설정 (Developer settings).

  • Mac: keep the server URL http://localhost:8010/v1 and tool_choice auto. Set the model name to your --served-model-name, for example KORMo-VL/KORMo-Web-4B. The default model name KORMo-VL/KORMo-Web-4B-mlx-4bit already matches the 4-bit build.
  • NVIDIA GPU: set the server URL to http://<server>:8000/v1 (or localhost), the model name to KORMo-Web-4B, and tool_choice to none.

4. Discord (optional). Run node relay/discord-relay.mjs in the cloned folder and follow the prompts to send tasks to the agent by mentioning a bot in a channel.

Full step-by-step instructions (Korean) are in the extension README.

Quickstart

Tested with vLLM 0.23.0 and transformers 5.12.1. The model runs on a single GPU (TP=1).

Run the agent

1. Install and download

pip install -U "vllm>=0.23.0" openai playwright huggingface_hub
playwright install chromium
hf download KORMo-VL/KORMo-Web-4B --local-dir KORMo-Web-4B

2. Start the model server

vllm serve ./KORMo-Web-4B --served-model-name KORMo-Web-4B \
  --max-model-len 131072 --limit-mm-per-prompt '{"image":1,"video":0}'

3. Run a browser task

python KORMo-Web-4B/examples/agent_playwright.py --model KORMo-Web-4B \
  --start-url https://www.danawa.com/ \
  --task "다나와에서 50만 원 이하, 10인치 이상 태블릿 3개를 찾아서 가격을 알려줘"

The agent drives a headless Chromium, prints each action as it goes, and finishes with the model's answer. Add --headful to watch the browser window.

On a Mac (vllm-mlx, see How to Start above), point the script at that server and pass --tool-choice auto:

python KORMo-Web-4B/examples/agent_playwright.py --endpoint http://localhost:8010/v1 \
  --model KORMo-VL/KORMo-Web-4B --tool-choice auto --headful \
  --start-url https://www.danawa.com/ \
  --task "다나와에서 50만 원 이하, 10인치 이상 태블릿 3개를 찾아서 가격을 알려줘"

Inference example (one step)

Send one screenshot with the request and get back one action:

import base64, json
from openai import OpenAI

SYSTEM = open("KORMo-Web-4B/system_prompt.md").read().strip()
TOOLS = json.load(open("KORMo-Web-4B/tools.json"))
shot = base64.b64encode(open("screen.png", "rb").read()).decode()

obs = ("<observation>\nscreen size: 1280 x 720\nCurrent tab: 0\nAvailable tabs:\n"
       "- Tab 0 (active): https://www.google.com/\n\nurl: https://www.google.com/\nscreenshot:\n")
messages = [
    {"role": "system", "content": SYSTEM},
    {"role": "user", "content": [
        {"type": "text", "text": "<user_request>\n구글에서 KAIST 를 검색해서 공식 홈페이지를 열어줘.\n</user_request>\n" + obs},
        {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{shot}"}},
        {"type": "text", "text": "\n</observation>"},
    ]},
]
client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")
out = client.chat.completions.create(
    model="KORMo-Web-4B", messages=messages, tools=TOOLS,
    tool_choice="none",  # vLLM. On vllm-mlx send "auto" and read message.tool_calls (see below)
    temperature=0.6, top_p=0.95, max_tokens=16384,
    extra_body={"top_k": 20, "chat_template_kwargs": {"enable_thinking": True}},
)
print(out.choices[0].message.content)

The output looks like this (illustrative):

<think>
I can see the Google home page with an empty search box in the center. I have not searched yet.
My plan: 1. Click the search box. 2. Type "KAIST". 3. Press Enter and open the official site.
</think>

<tool_call>
<function=click>
<parameter=x>
500
</parameter>
<parameter=y>
427
</parameter>
</function>
</tool_call>

tool_choice depends on the server. On vLLM, send "none": the tool schema is still rendered into the prompt and the call comes back as text in message.content, as above. Sending tools without tool_choice makes vLLM return HTTP 400 unless the server was started with --enable-auto-tool-choice. On vllm-mlx, "none" drops the tool schema from the prompt, so send "auto"; the call then arrives in message.tool_calls. examples/agent_playwright.py turns it back into the <tool_call> text above (--tool-choice auto).

Coordinates in the tool call are normalized to 0–1000. Convert them to pixels with x / 1000 × width and y / 1000 × height.

How It Works

user       <user_request> task </user_request>  + observation (tabs, url, screenshot)
assistant  <think> what I see · what is done · plan </think>  <tool_call> one action </tool_call>
tool       action result  + new observation (newest screenshot only)
…          repeat until finished(answer)
  • Action space: 22 browser tools: click, type, hotkey, scroll, drag, goto, tab control, finished, and others (tools.json).
  • Memory: the full text history of thoughts, actions and observations is kept. Only the current screenshot is sent.
  • Recommended settings: temperature 0.6, top_p 0.95, top_k 20, thinking on, up to 16k tokens per step and 50 steps. These are already set in generation_config.json.

The exact message format, parsing rules and feedback strings are in HARNESS.md. Follow it if you build your own harness.

Evaluation Details

Tasks Online-Mind2Web: 300 English, 312 Korean
Judge GPT-6

Files

File Description
chat_template.jinja Chat template (Qwen3.5 XML tool calls, thinking on)
system_prompt.md · tools.json System prompt and 22-tool schema, verbatim from training
HARNESS.md Full harness specification
examples/agent_playwright.py Reference browser agent (Playwright + OpenAI-compatible API). --tool-choice none for vLLM, auto for vllm-mlx; --headful shows the browser. It does not execute upload.
config.json · generation_config.json · processor_config.json · tokenizer*.json Model, sampling and processor configs

한국어 요약

KORMo-Web-4B는 사람처럼 웹 화면을 보고 클릭·입력·탐색을 수행하는 4B 규모의 한영 시각 기반 웹 에이전트입니다.

  • 대규모 웹 사전학습: 1,000만 건 이상의 한영 웹 학습 샘플로 UI 요소의 위치·의미·구조·기능 등을 세밀하게 학습(Atomic-Level Pretraining)
  • 실제 웹 행동 학습: 10만 건 이상의 한국인 한영 웹 사용 궤적을 학습해 여러 웹사이트와 탭을 넘나드는 복잡한 작업 수행
  • 뛰어난 성능: Online-Mind2Web에서 영어·한국어 모두 42.3% 성공률, 비교한 공개 모델 중 최고 성능(기존 최고 대비 +9.6%p / +22.1%p)
  • 시각 중심 웹 조작: 별도의 DOM 정보 없이 스크린샷만으로 웹을 이해하고, 행동 결과를 확인하며 다음 작업 수행
  • 모델·데이터 공개: Apache-2.0 모델 가중치, 대규모 학습 데이터, 웹 사용 궤적, 평가 벤치마크 및 실행 코드 공개

Acknowledgements

Built on Qwen3.5-4B. Evaluated on Online-Mind2Web and its Korean counterpart.

Contact

  • KyungTae Lim, Professor at KAIST. ktlim@kaist.ac.kr

Contributor

Downloads last month
374
Safetensors
Model size
5B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for KORMo-VL/KORMo-Web-4B

Finetuned
Qwen/Qwen3.5-4B
Finetuned
(1,523)
this model
Quantizations
2 models

Spaces using KORMo-VL/KORMo-Web-4B 2

Collection including KORMo-VL/KORMo-Web-4B