---

license: apache-2.0

base_model:

• mistralai/Ministral-3-14B-Instruct-2512-BF16

---

This is quantized version of Aratako/Ministral-3-14B-Instruct-2512-BF16-TextOnly created using llama.cpp

This model is the text-only component extracted from the Vision-Language Model mistralai/Ministral-3-14B-Instruct-2512-BF16.

Usage

You can load this model using AutoModelForCausalLM as shown below:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Aratako/Ministral-3-14B-Instruct-2512-TextOnly-BF16"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="cuda",
)

messages = [
    {
        "role": "user",
        "content": "Tell me a joke about computers.",
    },
]

input_ids = tokenizer.apply_chat_template(messages, return_tensors="pt").to("cuda")

output = model.generate(
    input_ids, max_new_tokens=512, pad_token_id=tokenizer.eos_token_id
)

decoded_output = tokenizer.decode(
    output[0][len(input_ids[0]) :], skip_special_tokens=True
)
print(decoded_output)

Original Model Information

This is a weight extraction of the original VLM. For detailed benchmarks, licensing details, and architectural information, please refer to the original model card: mistralai/Ministral-3-14B-Instruct-2512-BF16