whatmodel.sh / Ministral 3 14B Reasoning 2512

Ministral 3 14B Reasoning 2512

+ contribute

Ministral 3 14B Reasoning 2512 is an open-weights language model from Mistral, released December 2025. Its published size is 13.9B parameters and a 262k-token context window. Licence: apache-2.0.

Model information sources

· Parameter count ↗

Sum of the two components the card states (13.5B LM + 0.4B vision encoder = 13.9B). The repo's safetensors header gives 13,945,031,680 parameters (HF API); the sibling non-reasoning catalogue entry uses 13.9. Dense model: active_b = params_b.

57.5±1.0#228 of 719 LLMsfrom 20 benchmark results across 6 sources

$ how it compares

The LLMs ranked either side of Ministral 3 14B Reasoning 2512, out of 828 with a score, then more from Mistral.

#	model	score	where	RAM	released
270	Llama 3.3 Nemotron Super 49B v1.5 NVIDIA	58±1	local	~18GB	2025-07
271	Nemotron 3 Nano 30B-A3B NVIDIA	58±1	api $0.087/M +local	~20GB	2025-12
272	Ministral 3 14B Reasoning 2512 Mistral	58±1	local	~10GB	2025-12
273	GPT-4o (2024-08-06) OpenAI	57±1	api $4.38/M		2024-08
274	Gemini 2.0 Flash Thinking Experimental (01-21) Google	57±2	comparison only		2025-01
	more from Mistral
79	Mistral Large 4 (public preview) Mistral	71±2	api $2.06/M		2026-10
174	Mistral Medium 3.5 (128B) Mistral	65±1	api $3.00/M +local	~42GB	2026-04
180	Mistral Medium 3.1 (2508) Mistral	64±2	comparison only		2025-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #237 of 300 * · 3.0% of its score · 2 tests

Beats o3-mini here, though they rank higher overall. Trails Granite 4.0 1B, SmolLM3 3B, Phi-3.5-mini-instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL V4 (December 2025 archive)	35.9% ✎	▁▁▁▂▇▆▆▆█▂▄▂ beats 43/111	hard · avg 33%	▇ 2.1%	at its level
BFCL v3	52.2% ✎	▁▁▁▁▁▃▅█▆▅▄▁ beats 50/170	medium · avg 58%	▇ 0.9%	blunt · at its level

coding▁▁▁▃▆██▇▆▄▃▁ #224 of 458 · 11% of its score · 2 tests

Beats Qwen3.5 4B, Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B) here, though they rank higher overall. Trails QwQ-32B-Preview, Ministral 3 3B Reasoning 2512, GPT-4 (0613) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LiveCodeBench (version unspecified) lab +2	64.6% ✎	▁···▁▁▃▅▅█▄▅ beats 29/37	medium · avg 47%	▇▇▇ 7.8%	at its level · small run · several runs
LiveCodeBench v6 (window unspecified)	42.7% ✎	▂▃▄▄▅██▅▄▇▃▆ beats 12/65	medium · avg 43%	▇ 2.9%	at its level

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #302 of 393 * · 0.7% of its score · 1 test

Trails LFM2 350M, Gemma 3 1B, Flan-T5 XXL here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
IFEval (aggregation unspecified)	59.7% ✎	▁▁▃▃▂▄▄▆█▃▂▁ beats 21/141	easy · avg 76%	▇ 0.7%	noisy

math█·····█····█ #1 of 3 · 28% of its score · 2 tests

test	result	vs the others in its run	difficulty	counts	why
AIME 2025 lab +2	85.0% ✎	▁·▁▁▁▂▅▆█▇▅▁ beats 99/133	medium · avg 64%	▇▇▇▇▇▇ 14%	sharp · small run · several runs
AIME 2024 lab +2	89.8% ✎	▁▁▁▂▁▃▁▅███▃ beats 72/85	easy · avg 73%	▇▇▇▇▇ 14%	sharp · small run · several runs

other tests8 not in a task · 57% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA Diamond +3	46.0% ✎	▁▁▁▁▂▄▅▅███▄ beats 158/310	hard · avg 51%	▇▇▇▇▇▇▇▇▇ 23%	sharp · at its level · small run · several runs
Medmarks-V mean win rate (19 verifiable medical benchmarks, 71 configurations)	0.4807 ✎	▁▄▃▃▇▄█▃▅▁▁▁ beats 14/30	avg 0.423	▇▇▇▇▇ 13%	sharp
Medmarks-V weighted mean win rate (61 models, 71 configurations)	48.1% ✎	▂▃▅▂▆▃█▅█▅▃▆ beats 13/34	medium · avg 42%	▇▇▇ 6.3%	blunt · at its level
EuroEval Life in the UK (English knowledge MC), validation split, 5-shot, accuracy	93.2% ✎	▂·▂▂·▂·▃▃·█▃ beats 10/14	easy · avg 80%	▇▇ 5.0%	typical
EuroEval HellaSwag (English), validation split, 5-shot generative MC, accuracy	86.9% ✎	▂··▂▂·▂▆▆█·▄ beats 14/15	medium · avg 63%	▇ 3.4%	at its level
MMLU-Redux	84.4% ✎	▁▁▃▁▂▃▄▅▇▅█▄ beats 17/63	easy · avg 77%	▇ 3.2%	typical
SurgiQ surgical MCQ (zero-shot, log-probability scoring)	63.5% ✎	▂···▂▂·▃▂▃██ beats 13/17	medium · avg 53%	▇ 2.0%	at its level
GSM-Plus	86.5% ✎	▁··▁▁▁·▃▁▄█▁ beats 8/16	easy · avg 83%	▇ 1.4%	blunt

$ where to run it

llama.cpp · recommended

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 ↗
Run Ministral 3 14B Reasoning 2512 · needs 8.24 GB · from huggingface.co/mistralai
chat in the terminal
llama cli -hf mistralai/Ministral-3-14B-Reasoning-2512-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf mistralai/Ministral-3-14B-Reasoning-2512-GGUF:Q4_K_M

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run Ministral 3 14B Reasoning 2512 · needs 8.24 GB · from huggingface.co/mistralai
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/mistralai/Ministral-3-14B-Reasoning-2512-GGUF@Q4_K_M -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run Ministral 3 14B Reasoning 2512 · needs 8.24 GB · from huggingface.co/mistralai
chat in the terminal
ollama run hf.co/mistralai/Ministral-3-14B-Reasoning-2512-GGUF:Q4_K_M

MLX (Mac)

Install MLX (Mac) · Apple silicon Macs only, text only.
with uv
uv tool install mlx-lm
or
pip install mlx-lm
Run Ministral 3 14B Reasoning 2512 · needs ~8.4 GB, Mac only · from huggingface.co/hunterbown
chat in the terminal
mlx_lm.chat --model hunterbown/Ministral-3-14B-Reasoning-2512-MLX-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model hunterbown/Ministral-3-14B-Reasoning-2512-MLX-4bit

Weights and documentation: huggingface.co/mistralai ↗

$ how the score was worked out · mistralai/Ministral-3-8B-Reas… 27%, mistralai/Ministral-3-3B-Reas… 27%, Medmarks: A Comprehensive Ope… 19% and 3 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

source	results	share of the score	licence
mistralai/Ministral-3-8B-Reasoning-2512 model card · lab-reported	4	▇▇▇▇▇▇▇▇ 27%	Apache-2.0
mistralai/Ministral-3-3B-Reasoning-2512 model card · lab-reported	4	▇▇▇▇▇▇▇▇ 27%	Apache-2.0
Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks — authors' own runs of 61 models	2	▇▇▇▇▇▇ 19%	CC BY 4.0
JetBrains/Mellum2-12B-A2.5B-Thinking model card	7	▇▇▇▇▇ 15%	Apache-2.0
EuroEval leaderboards (results collection), English generative leaderboard	2	▇▇▇ 8%	MIT
SurgiQ: A Large-Scale Multi-Domain Benchmark for Evaluating Surgical Understanding in Large Language Models (arXiv:2606.08071v1)	1	▇ 2%	CC BY 4.0
prior · pull toward the middle		▇ 0%
weighted average		= 57.5

not counted (4)

mistralai/Ministral-3-14B-Reasoning-2512 model card	AIME 2025 [correct]	85.0%	the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts
mistralai/Ministral-3-14B-Reasoning-2512 model card	AIME 2024 [correct]	89.8%	the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts
mistralai/Ministral-3-14B-Reasoning-2512 model card	GPQA Diamond [correct]	71.2%	the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts
mistralai/Ministral-3-14B-Reasoning-2512 model card	LiveCodeBench (version unspecified) [correct]	64.6%	the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit 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 2026-10-11 02:20 UTC · sources and licences · json