whatmodel.sh / Arch-Agent-3B

Arch-Agent-3B

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

Arch-Agent-3B is an open-weights language model from Katanemo, released June 2025. Its published size is 3.09B parameters. Licence: other (katanemo-research).

Catalogue identity notes

Maker repo katanemo/Arch-Agent-3B @46927c9: licence tag other (katanemo-research). 3,085,938,688 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2025-06: month the maker's Hugging Face repo was created (createdAt 2025-06-20).

52.2±4.7#392 of 949 LLMsfrom 1 benchmark result across 1 source* thinly tested, treat the rank as provisional

$ how it compares

The LLMs ranked either side of Arch-Agent-3B, out of 1132 with a score, then more from Katanemo.

#modelscorewhereRAMreleased
508Qwen2.5-Math 7B Instruct Alibaba52±2local~6GB2024-09
509LFM2.5 8B-A1B Liquid AI52±1local~6.5GB2026-05
510Arch-Agent-3B Katanemo52±5local~3.3GB2025-06
511GPT-4 (0314) OpenAI52±1comparison only2023-03
512GLM-4 (0520) Z.ai52±1comparison only2024-05
more from Katanemo
328Arch-Agent-32B Katanemo58±5local~21GB2025-06
570Arch-Agent-1.5B Katanemo50±5local~2.4GB2025-06

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▄▄▃▂▁ #227 of 334 * · 90% of its score · 1 test

Beats o1 (2024-12-17), o3-mini, DeepSeek R1 (Jan '25) here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Granite 4.1 3B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BFCL V4 (December 2025 archive)35.4% ✎▁▁▁▂██▅▅▆▂▂▂ beats 38/77hard · avg 32%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 90%at its level

$ where to run it

llama.cpp · recommended

version

  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 Arch-Agent-3B · needs ~1.9 GB · from huggingface.co/mradermacher
    chat in the terminal
    llama cli -hf mradermacher/Arch-Agent-3B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf mradermacher/Arch-Agent-3B-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 Arch-Agent-3B · needs ~1.9 GB · from huggingface.co/mradermacher

    open in LM Studio ↗

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

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Arch-Agent-3B · needs ~1.9 GB · from huggingface.co/mradermacher
    chat in the terminal
    ollama run hf.co/mradermacher/Arch-Agent-3B-GGUF:Q4_K_M

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 Arch-Agent-3B · needs ~1.7 GB, Mac only · from huggingface.co/huangang
    chat in the terminal
    mlx_lm.chat --model huangang/Arch-Agent-3B-mlx-4Bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model huangang/Arch-Agent-3B-mlx-4Bit

Other ways to run it

  1. Run it as published · needs ~6.17 GB · from huggingface.co/katanemo (official)
    uv run --with torch --with 'transformers<5' --with accelerate python -c "from transformers import pipeline; print(pipeline('text-generation', model='katanemo/Arch-Agent-3B', torch_dtype='auto', device_map='auto')('Once upon a time', max_new_tokens=40)[0]['generated_text'])"

Official weights and documentation: huggingface.co/katanemo ↗

$ how the score was worked out · Berkeley Function Calling Lea… 90%

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
Berkeley Function Calling Leaderboard result archive1▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 90%Apache-2.0
prior · pull toward the middle▇▇▇ 10%
weighted average= 52.2

$ 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