whatmodel.sh / QED-Nano

QED-Nano

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

QED-Nano is an open-weights language model from LM Provers, released February 2026. Its published size is 4.02B parameters and a 262k-token context window. Licence: apache-2.0.

Catalogue identity notes

Proposed so the three MathArena ArXivMath runs (Dec 2025, Jan 2026, Feb 2026; 18-24 catalogue models each) that include it count it. LM Provers' 4B proof-writing model, post-trained (SFT + RL with a reasoning cache, arXiv 2602.03773) from Qwen/Qwen3-4B-Thinking-2507; not the intermediate lm-provers/QED-Nano-SFT. MathArena config configs/models/other/qed-nano.yaml (eth-sri/matharena b89f2f0ad64c): model 'lm-provers/QED-Nano', huggingface_id lm-provers/QED-Nano, creator LM-Provers, parameters 4, date 2026-02-15. Repo created 2026-02-12 (HF API createdAt). The card documents 'vllm serve lm-provers/QED-Nano'; gb = the two BF16 safetensors shards at the pinned revision (4,967,215,360 + 3,077,766,632 bytes), weights only.

Model information sources

Release date ↗ · Weights licence ↗ · Parameter count ↗

Hugging Face safetensors metadata at the pinned revision: BF16 4,022,468,096 parameters (Qwen3ForCausalLM, 36 layers); the card calls it a 4B model.

57.0±2.3#237 of 719 LLMsfrom 3 benchmark results across 1 source

$ how it compares

The LLMs ranked either side of QED-Nano, out of 828 with a score.

#modelscorewhereRAMreleased
281Qwen2.5-Omni 7B Alibaba57±2local~6.4GB2025-03
282GPT-4o (2024-05-13) OpenAI57±0api $7.50/M2024-05
283QED-Nano LM Provers57±2local~4.1GB2026-02
284Llama 4 Maverick (402B-A17B) Meta57±1api $0.42/M +local~65GB2025-04
285Claude 3.5 Sonnet (2024-06-20) Anthropic57±1comparison only2024-06

see all LLMs · find the ones that fit your machine

$ what it's good at

math▁▂▃▅▆▅▄█▄▁▁▁ #183 of 398 · 98% of its score · 3 tests

Beats MiniMax-M2.1, DeepSeek V3 0324, Qwen3 235B-A22B (Apr 2025, hybrid) here, though they rank higher overall. Trails Phi-4-mini-reasoning 3.8B, OpenThinker3 7B, Olmo 3 7B Think here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
MathArena ArXivMath January 2026 (23 problems)37.0% ✎▂··▄▂▄█▆▆▄▄█ beats 1/23hard · avg 22%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 40%at its level
MathArena ArXivMath February 2026 (32 problems)14.1% ✎▃▁▁▁█▇▄▃▁▃·▃ beats 0/24very hard · avg 9.1%▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 34%typical
MathArena ArXivMath December 2025 (17 problems)26.5% ✎▃·▃▃█▁·▄·▁▁▁ beats 0/18hard · avg 19%▇▇▇▇▇▇▇▇▇ 23%at its level

$ 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 QED-Nano · needs ~2.5 GB · from huggingface.co/ijohn07
    chat in the terminal
    llama cli -hf ijohn07/QED-Nano-Q4_K_M-GGUF:q4_k_m
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf ijohn07/QED-Nano-Q4_K_M-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 QED-Nano · needs ~2.5 GB · from huggingface.co/ijohn07

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/ijohn07/QED-Nano-Q4_K_M-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 QED-Nano · needs ~2.5 GB · from huggingface.co/ijohn07
    chat in the terminal
    ollama run hf.co/ijohn07/QED-Nano-Q4_K_M-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 QED-Nano · needs ~4.3 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/QED-Nano-8bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/QED-Nano-8bit

Other ways to run it

  1. Run it as published · needs ~8.04 GB, NVIDIA only · from huggingface.co/lm-provers (official)
    vllm serve lm-provers/QED-Nano

Official weights and documentation: huggingface.co/lm-provers ↗

$ how the score was worked out · MathArena ArXivMath / BrokenA… 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

sourceresultsshare of the scorelicence
MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets3▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 98%CC BY-SA 4.0
prior · pull toward the middle▇ 2%
weighted average= 57.0

$ 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