whatmodel.sh / Ministral 8B (2410)

Ministral 8B (2410)

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

Ministral 8B (2410) is an open-weights language model from Mistral, released October 2024. Its published size is 8.02B parameters and a 128k-token context window. Licence: mistral-research.

Catalogue identity notes

Comparison model from the Epoch AI internal-evaluation CSVs (Epoch release date 2024-10-16). API id ministral-8b-2410 ('Ministral 8B', 24.10; retired 12/31/2025); weights Ministral-8B-Instruct-2410 under the Mistral Research License (non-commercial). Distinct from Ministral 3 8B (2512). 8,019,808,256 parameters; command from the model card; size is consolidated.safetensors (16.04 GB).

43.7±0.7#471 of 719 LLMsfrom 39 benchmark results across 12 sources

$ how it compares

The LLMs ranked either side of Ministral 8B (2410), out of 828 with a score, then more from Mistral.

#modelscorewhereRAMreleased
558Qwen3 1.7B Alibaba44±1local~2.6GB2025-04
559Phi-4-mini 3.8B Microsoft44±1api $0.13/M +local~3.8GB2025-02
560Ministral 8B (2410) Mistral44±1local~6.2GB2024-10
561Granite 3.2 8B Instruct IBM44±1local~6.6GB2025-02
562Code Llama 70B Python Meta44±1local~28GB2024-01
more from Mistral
79Mistral Large 4 (public preview) Mistral71±2api $2.06/M2026-10
174Mistral Medium 3.5 (128B) Mistral65±1api $3.00/M +local~42GB2026-04
180Mistral Medium 3.1 (2508) Mistral64±2comparison only2025-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #253 of 300 * · 1.8% of its score · 3 tests

Beats DeepSeek R1 0528 Qwen3 8B, Magistral Small 1.0 (2506), Llama 4 Scout (109B-A17B) here, though they rank higher overall. Trails Granite 4.0 H 1B, Granite 4.0 1B, SmolLM3 3B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Galileo Agent Leaderboard v1 (tool selection quality)68.9% ✎▃▃▃▃▆▅▆▅█▂▅▂ beats 3/31easy · avg 79%▇ 0.9%blunt · at its level
BFCL v351.5% ✎▁▁▁▁▁▃▅█▅▄▁▁ beats 41/114medium · avg 52%▇ 0.9%blunt · at its level
BFCL V4 (December 2025 archive)11.1% ✎▁▁▁▂▇▆▆▆█▂▄▂ beats 0/111hard · avg 32%—failed run: Same pattern as Nemotron Ultra: zero on every category that needs a call and 100% on irrelevance, so the FC handler never got a parseable call.
coding█··········█ #1 of 2 * · 3.1% of its score · 2 tests
testresultvs the others in its rundifficultycountswhy
HumanEval lab76.8% ✎▁▁▁▁▁▃▅▅█▃▂▁ beats 148/190easy · avg 83%▇ 1.7%sharp · at its level · small run
MBPP lab70.0% ✎▁▁▁▄▅█▄▅▃▁▂▁ beats 64/81easy · avg 76%▇ 1.4%at its level · small run
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #284 of 393 · 8.3% of its score · 4 tests

Beats Ministral 3 14B Reasoning 2512, Gemma 4 E4B, Ministral 3 8B here, though they rank higher overall. Trails Flan-T5 XXL, LFM2 700M, EXAONE 3.5 2.4B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1237.6 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 88/379avg 1268▇▇ 3.9%blunt
AlpacaEval 2 length-controlled31.4% ✎▃▃·▁▁█▇▅·▁·▁ beats 8/22medium · avg 42%▇ 2.2%at its level
Arena-Hard (version/judge unspecified) lab70.9% ✎▃▃▂▅▃▆▅█▅▅▂▃ beats 21/30easy · avg 70%▇ 1.3%at its level · small run
IFEval prompt-level loose56.4% ✎▁▃▃▃▁▄▁▁█▃▃▁ beats 3/22easy · avg 80%▇ 0.9%at its level
math▁▂▃▅▆▅▄█▄▁▁▁ #331 of 398 · 24% of its score · 5 tests

Beats GPT-3.5 Turbo (0125), Qwen2-VL 7B Instruct, DBRX Instruct here, though they rank higher overall. Trails Qwen3 0.6B, Gemma 7B (base), DeepSeek R1 Distill Qwen 1.5B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
GSM8K +180.0% ✎▂▃··▃▂▅█▆▃▂▃ beats 7/22very easy · avg 91%▇▇▇▇ 9.4%sharp · several runs
MGSM56.7% ✎▅▃▃█·▅···▃·▃ beats 5/10easy · avg 71%▇▇ 4.7%at its level
MATH level 514.9% ✎▂▂▅▆▄█▆▃▅▂▃▄ beats 13/92hard · avg 32%▇▇ 4.2%sharp
MATH41.6% ✎█▄██···▄▄▄·▄ beats 0/10medium · avg 64%▇ 2.9%at its level
MATH (subset unspecified) +140.0% ✎▃▃█▂▂▆·····▁ beats 7/22medium · avg 61%▇ 2.6%at its level · several runs
other tests21 not in a task · 63% of its score
testresultvs the others in its rundifficultycountswhy
MMLU +168.5% ✎▁▂·▁·▁▁█·▃▃▅ beats 8/22easy · avg 77%▇▇▇▇▇ 11%at its level · several runs
BIG-Bench Hard +170.8% ✎▁▄·▁▁·▅█·▄▄▁ beats 13/22easy · avg 75%▇▇▇▇ 11%sharp · at its level · several runs
ARC Challenge82.7% ✎▂▂▂█▂···▄··▂ beats 4/10very easy · avg 91%▇▇▇▇ 9.0%sharp
OpenBookQA80.2% ✎▂··▄█▂·▂···▄ beats 5/10easy · avg 87%▇▇ 5.2%at its level
MMLU-Pro36.6% ✎█▄▄·█▄▄▄·▄·▄ beats 1/10medium · avg 57%▇ 3.6%at its level
Multilingual MMLU (Microsoft)53.7% ✎█▄▄██··█···▄ beats 4/10easy · avg 65%▇ 3.3%at its level
GPQA Diamond27.1% ✎▁▁▃▁▂▄▆▄▅▇█▅ beats 9/208hard · avg 40%▇ 2.5%sharp
WinoGrande63.1% ✎▃····▃▃█▃·▅▅ beats 2/10medium · avg 74%▇ 2.2%at its level
13 more tests
testresultvs the others in its rundifficultycountswhy
HellaSwag73.5% ✎▃▅█·▃▃▃···▃▃ beats 3/10easy · avg 84%▇ 2.0%at its level
BoolQ80.5% ✎▃··▃··██▃·▃▃ beats 6/10easy · avg 85%▇ 1.9%at its level
Social IQA77.6% ✎▃·▃▅▅█·▃···▃ beats 9/10easy · avg 78%▇ 1.8%blunt · at its level
BRIDGE clinical text, zero-shot average (87 tasks)30.4% ✎▁▁▁▁▁▂▂▃▄▆█▃ beats 32/97hard · avg 32%▇ 1.7%blunt · at its level
TruthfulQA MC2 +155.5% ✎▃▃▂·▃▃·▅█▂▅▂ beats 7/21medium · avg 63%▇ 1.4%blunt · at its level · several runs
AraGen-v3 3C3H score (Inception, Claude Sonnet 4 judge)17.6% ✎▁▁·▁▂▂▄█▄▅▄▄ beats 15/66hard · avg 31%▇ 1.4%typical
GPQA (subset unspecified)26.3% ✎█··▄▂··▂▂··▄ beats 5/10very hard · avg 33%▇ 0.9%sharp · too hard for it · noisy
PopQA20.2% ✎▁▁▁▃▃▃▁█▁▃·▄ beats 6/21hard · avg 32%▇ 0.9%blunt
Open PT-LLM Leaderboard average (9 Portuguese tasks)71.3% ✎▃▁▁▁▁▁▁▁▂█▅▁ beats 181/276medium · avg 73%▇ 0.9%at its level · noisy
PIQA81.2% ✎▃·▅·▃█·▃▅··▃ beats 7/10easy · avg 83%▇ 0.6%blunt · at its level
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy82.5% ✎▁▁▁▁▁▁▁▂▄█▄▃ beats 79/146easy · avg 83%▇ 0.6%at its level · noisy
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)92.5% ✎▁▁▂▃▄▄█▆▃▂▂▁ beats 39/140very easy · avg 94%▇ 0.3%blunt · noisy
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)92.6% ✎▁▂▁▂▅█▂▃▄▃▁▁ beats 65/98very easy · avg 90%▇ 0.0%blunt

$ 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 Ministral 8B (2410) · needs ~4.9 GB · from huggingface.co/bartowski
    chat in the terminal
    llama cli -hf bartowski/Ministral-8B-Instruct-2410-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf bartowski/Ministral-8B-Instruct-2410-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 Ministral 8B (2410) · needs ~4.9 GB · from huggingface.co/bartowski

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/bartowski/Ministral-8B-Instruct-2410-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 Ministral 8B (2410) · needs ~4.9 GB · from huggingface.co/bartowski
    chat in the terminal
    ollama run hf.co/bartowski/Ministral-8B-Instruct-2410-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 Ministral 8B (2410) · needs ~4.5 GB, Mac only · from huggingface.co/alexgusevski
    chat in the terminal
    mlx_lm.chat --model alexgusevski/Ministral-8B-Instruct-2410-q4-mlx
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model alexgusevski/Ministral-8B-Instruct-2410-q4-mlx

Other ways to run it

  1. Run it as published · needs ~19.75 GB, NVIDIA only · from huggingface.co/mistralai
    vllm serve mistralai/Ministral-8B-Instruct-2410 --tokenizer_mode mistral --config_format mistral --load_format mistral

Weights and documentation: huggingface.co/mistralai ↗

$ how the score was worked out · microsoft/Phi-4-mini-instruct… 56%, 2 OLMo 2 Furious (Team OLMo,… 21%, Epoch AI — internal evaluatio… 7% and 9 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
microsoft/Phi-4-mini-instruct model card16▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 56%MIT
2 OLMo 2 Furious (Team OLMo, Ai2) — authors' own evaluations of OLMo 2 Instruct and peer instruct models8▇▇▇▇▇▇ 21%CC BY 4.0
Epoch AI — internal evaluations2▇▇ 7%CC BY 4.0
Mistral AI vendor-reported benchmark scores for its own models (Hugging Face model cards) · lab-reported4▇▇ 6%Factual vendor-reported score (not licensed content)
LMArena leaderboard dataset1▇ 4%CC BY 4.0
BRIDGE Medical Leaderboard (Harvard Medical School / Brigham and Women's YLab)1▇ 2%MIT
Inception Arabic/Hindi leaderboards (AraGen-v3, HindiGen-v1, Arabic IFEval)1▇ 1%Apache-2.0
Galileo Agent Leaderboard (v1 and v2 results)1▇ 1%Apache-2.0
Open Portuguese LLM Leaderboard (archived)1▇ 1%Apache-2.0
Berkeley Function Calling Leaderboard result archive1▇ 1%Apache-2.0
EuroEval leaderboards (results collection), English generative leaderboard1▇ 1%MIT
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)2▇ 0%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 43.7

not counted (1)

Berkeley Function Calling Leaderboard result archiveBFCL V4 (December 2025 archive) [correct]11.1%failed run: Same pattern as Nemotron Ultra: zero on every category that needs a call and 100% on irrelevance, so the FC handler never got a parseable call.

$ 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