whatmodel.sh / Magicoder-S-DS 6.7B

Magicoder-S-DS 6.7B

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

Magicoder-S-DS 6.7B is an open-weights language model from UIUC (ise-uiuc), released December 2023. Its published size is 6.74B parameters. Licence: deepseek.

Catalogue identity notes

Comparison-only. Maker repo ise-uiuc/Magicoder-S-DS-6.7B @b3ed7cb: licence tag other (deepseek). 6,740,512,768 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2023-12: month the maker's Hugging Face repo was created (createdAt 2023-12-03). OSS-Instruct + Evol-Instruct fine-tune of DeepSeek-Coder 6.7B base; not Magicoder-DS or the CodeLlama-based -CL variants.

48.5±0.7#372 of 719 LLMsfrom 10 benchmark results across 4 sources

$ how it compares

The LLMs ranked either side of Magicoder-S-DS 6.7B, out of 828 with a score, then more from UIUC (ise-uiuc).

#modelscorewhereRAMreleased
443Yi-Coder 9B Chat 01.AI49±1local~6.8GB2024-08
444Qwen2-Math 7B Instruct Alibaba49±2local~6GB2024-08
445Magicoder-S-DS 6.7B UIUC (ise-uiuc)49±1local~5.8GB2023-12
446InternLM2.5 7B Chat Shanghai AI Laboratory (InternLM)49±2local~6.1GB2024-06
447Hermes 3 Llama 3.1 70B Nous Research49±1local~44GB2024-07
more from UIUC (ise-uiuc)
482Magicoder-S-CL-7B UIUC (ise-uiuc)47±1local~5.8GB2023-12

see all LLMs · find the ones that fit your machine

$ what it's good at

coding▁▁▁▃▆██▇▆▄▃▁ #272 of 458 · 75% of its score · 5 tests

Beats Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B), GLM-4.7-Flash here, though they rank higher overall. Trails DeepSeek LLM 67B Chat, Mistral NeMo Instruct (2407), QwQ-32B-Preview here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
HumanEval +176.8% ✎▁·▁▁▃▄▆▅█▃▁▁ beats 78/99easy · avg 79%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 44%sharp · several runs
HumanEval+71.3% ✎▁·▁▁▁▃▄▄▄█▃▁ beats 74/99easy · avg 74%▇▇▇▇▇▇▇▇▇ 23%sharp · at its level
MBPP+69.0% ✎▁···▁▁▂▅▇█▆▂ beats 59/75easy · avg 68%▇▇ 4.5%at its level
CRUXEval-I41.7% ✎▁··▁▅▂█▇▁··▁ beats 12/22medium · avg 54%▇ 1.6%at its level
CRUXEval-O44.4% ✎▁·▂▃▁█▃▂···▁ beats 18/22medium · avg 53%▇ 1.5%blunt · at its level
other tests4 not in a task · 25% of its score
testresultvs the others in its rundifficultycountswhy
EvoEval (overall) lab50.0% ✎▂▂▃▄▄█▁▄▅▃▁▃ beats 35/42medium · avg 55%▇▇▇▇▇ 12%at its level
BigCodeBench (Instruct) lab36.0% ✎▃▂·▆▆██▆▅▃▃▃ beats 19/33medium · avg 41%▇▇ 5.0%at its level
BigCodeBench (Complete) lab47.5% ✎▂▁▁▄▂▅▇█▆▂▂▂ beats 46/58medium · avg 50%▇▇ 4.8%at its level
EvalPerf correctness (118 tasks, 20240328)69.8% ✎▂·▂▂▂▆▅▆▆▆▃█ beats 7/29easy · avg 76%▇ 2.6%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 Magicoder-S-DS 6.7B · needs ~4.1 GB · from huggingface.co/TheBloke
    chat in the terminal
    llama cli -hf TheBloke/Magicoder-S-DS-6.7B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf TheBloke/Magicoder-S-DS-6.7B-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 Magicoder-S-DS 6.7B · needs ~4.1 GB · from huggingface.co/TheBloke

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/TheBloke/Magicoder-S-DS-6.7B-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 Magicoder-S-DS 6.7B · needs ~4.1 GB · from huggingface.co/TheBloke
    chat in the terminal
    ollama run hf.co/TheBloke/Magicoder-S-DS-6.7B-GGUF:Q4_K_M

Other ways to run it

  1. Run it as published · needs ~26.96 GB · from huggingface.co/ise-uiuc (official)
    uv run --with torch --with 'transformers<5' --with accelerate python -c "from transformers import pipeline; print(pipeline('text-generation', model='ise-uiuc/Magicoder-S-DS-6.7B', torch_dtype='auto', device_map='auto')('Once upon a time', max_new_tokens=40)[0]['generated_text'])"

Official weights and documentation: huggingface.co/ise-uiuc ↗

$ how the score was worked out · EvalPlus / EvalPerf published… 59%, Top Leaderboard Ranking = Top… 28%, BigCodeBench paper (Zhuo et a… 10% and 1 more

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
EvalPlus / EvalPerf published evaluation runs4▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 59%Apache-2.0
Top Leaderboard Ranking = Top Coding Proficiency, Always? EvoEval: Evolving Coding Benchmarks via LLM — authors' own runs · lab-reported2▇▇▇▇▇▇▇▇ 28%CC BY 4.0
BigCodeBench paper (Zhuo et al., ICLR 2025), arXiv 2406.15877v4 · lab-reported2▇▇▇ 10%CC BY 4.0
CRUXEval published evaluation results (authors' generations and scores)2▇ 3%MIT
prior · pull toward the middle▇ 0%
weighted average= 48.5

$ 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