whatmodel.sh / Cogito v1 Preview Qwen 32B

Cogito v1 Preview Qwen 32B

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

Cogito v1 Preview Qwen 32B is an open-weights language model from Deep Cogito, released March 2025. Its published size is 32.76B parameters and a 131k-token context window. Licence: apache-2.0.

Catalogue identity notes

deepcogito/cogito-v1-preview-qwen-32B @118e6e4 (created 2025-03-31, licence apache-2.0). Deep Cogito v1 Preview 32B hybrid reasoning model (Mar 2025), sibling of the catalogued Llama 3B/8B previews. Independent rows: EuroEval English (both datasets).

Model information sources

Release date ↗ · Weights licence ↗

HF safetensors metadata 32,759,790,592 parameters (deepcogito/cogito-v1-preview-qwen-32B @118e6e4).

61.9±7.3#197 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 Cogito v1 Preview Qwen 32B, out of 1132 with a score, then more from Deep Cogito.

#modelscorewhereRAMreleased
253Qwen3-VL 32B Thinking Alibaba62±1api $0.28/M +local~22GB2025-10
254Gemini 2.0 Flash-Lite (001) Google62±2comparison only2025-02
255Cogito v1 Preview Qwen 32B Deep Cogito62±7local~21GB2025-03
256Llama-xLAM-2-8b-fc-r Salesforce62±4local~6.6GB2025-03
257Mistral Large 3 (675B-A41B) Mistral62±1api $0.75/M +local~108GB2025-12
more from Deep Cogito
266Cogito v2 Preview Llama 70B Deep Cogito61±7local~43GB2025-07
272Cogito v1 Preview Llama 70B Deep Cogito61±7local~43GB2025-03
285Cogito v1 Preview Qwen 14B Deep Cogito60±7local~10GB2025-03

see all LLMs · find the ones that fit your machine

$ what it's good at

other tests2 not in a task · 76% of its score
testresultvs the others in its rundifficultycountswhy
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy92.8% ✎▃·▁▁▂▁▂▅█▇▄▃ beats 98/100medium · avg 68%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 50%noisy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy93.0% ✎▁▁▁▁▁▁▁▂▄█▄▃ beats 128/146easy · avg 83%▇▇▇▇▇▇▇▇▇▇ 26%noisy

$ 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 Cogito v1 Preview Qwen 32B · needs ~19.8 GB · from huggingface.co/lmstudio-community
    chat in the terminal
    llama cli -hf lmstudio-community/cogito-v1-preview-qwen-32B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf lmstudio-community/cogito-v1-preview-qwen-32B-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 Cogito v1 Preview Qwen 32B · needs ~19.8 GB · from huggingface.co/lmstudio-community

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/lmstudio-community/cogito-v1-preview-qwen-32B-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 Cogito v1 Preview Qwen 32B · needs ~20 GB · from ollama.com
    chat in the terminal
    ollama run cogito:32b

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run Cogito v1 Preview Qwen 32B · needs ~18.4 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/deepcogito-cogito-v1-preview-qwen-32B-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/deepcogito-cogito-v1-preview-qwen-32B-4bit

Official weights and documentation: huggingface.co/deepcogito ↗

$ how the score was worked out · EuroEval leaderboards (result… 76%

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
EuroEval leaderboards (results collection), English generative leaderboard2▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 76%MIT
prior · pull toward the middle▇▇▇▇▇▇▇ 24%
weighted average= 61.9

$ 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