whatmodel.sh / Sky-T1 32B Preview

Sky-T1 32B Preview

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

Sky-T1 32B Preview is an open-weights language model from NovaSky (UC Berkeley), released January 2025. Its published size is 32.76B parameters. Licence: apache-2.0.

Catalogue identity notes

NovaSky-AI/Sky-T1-32B-Preview @1e3f4c6 (created 2025-01-07, licence apache-2.0). UC Berkeley NovaSky reasoning model (Jan 2025, Qwen2.5-32B-Instruct base). Independent rows: BFCL v3 (Prompt) and EuroEval English.

Model information sources

Release date ↗ · Weights licence ↗

HF safetensors metadata 32,763,876,352 parameters.

56.2±4.8#242 of 719 LLMsfrom 3 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of Sky-T1 32B Preview, out of 828 with a score.

#modelscorewhereRAMreleased
290Athene-V2-Chat 72B Nexusflow56±2local~44GB2024-11
291Gemini 2.0 Flash Experimental Google56±1comparison only2024-12
292Sky-T1 32B Preview NovaSky (UC Berkeley)56±5local~21GB2025-01
293Qwen Plus (2025-01-25) Alibaba56±1api $0.60/M2025-01
294Grok 2 (1212) SpaceXAI (xAI)56±1comparison only2024-12

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #244 of 300 * · 44% of its score · 1 test

Beats DeepSeek R1 0528 Qwen3 8B here, though they rank higher overall. Trails Granite 4.0 1B, SmolLM3 3B, Phi-3.5-mini-instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BFCL v352.4% ✎▁▁▁▁▁▃▅█▅▄▁▁ beats 46/114medium · avg 52%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 44%blunt · at its level
other tests2 not in a task · 46% of its score
testresultvs the others in its rundifficultycountswhy
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy91.0% ✎▃·▁▁▂▁▂▅█▇▄▁ beats 94/100medium · avg 68%▇▇▇▇▇▇▇▇▇▇▇▇ 30%at its level · noisy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy93.8% ✎▁▁▁▁▁▁▁▂▄█▄▃ beats 132/146easy · avg 83%▇▇▇▇▇▇▇ 16%noisy

$ where to run it

llama.cpp · recommended

  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 Sky-T1 32B Preview · needs ~19.85 GB · from huggingface.co/bartowski
    chat in the terminal
    llama cli -hf bartowski/Sky-T1-32B-Preview-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf bartowski/Sky-T1-32B-Preview-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 Sky-T1 32B Preview · needs ~19.85 GB · from huggingface.co/bartowski

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/bartowski/Sky-T1-32B-Preview-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 Sky-T1 32B Preview · needs ~19.85 GB · from huggingface.co/bartowski
    chat in the terminal
    ollama run hf.co/bartowski/Sky-T1-32B-Preview-GGUF:Q4_K_M

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 Sky-T1 32B Preview · needs ~18.4 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/Sky-T1-32B-Preview-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/Sky-T1-32B-Preview-4bit

Official weights and documentation: huggingface.co/NovaSky-AI ↗

$ how the score was worked out · EuroEval leaderboards (result… 46%, Berkeley Function Calling Lea… 44%

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▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 46%MIT
Berkeley Function Calling Leaderboard result archive1▇▇▇▇▇▇▇▇▇▇▇▇▇ 44%Apache-2.0
prior · pull toward the middle▇▇▇ 10%
weighted average= 56.2

$ 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