CompiwerAI/Mtrini-SVL-1.1-Merged
0679
๐ช Mtrini-SVL-1.1
The model is out of the lab.
Mtrini-SVL-1.1 is an open-weight ~8B model developed by Compiwer AI, based on Qwen3-VL-8B-Instruct.
It was trained with a focus on:
- ๐ง Reasoning
- ๐ป Coding
- ๐งฎ Mathematics
- ๐ฒ๐ฆ Moroccan Darija
- ๐ Multilingual tasks
And yes...
It's a Mtrini. ๐ช
๐ Self-Verifying Loop
SVL = Self-Verifying Loop
Our development concept:
THINK
โ
CHECK
โ
IMPROVE
โ
REPEAT ๐
The goal is to encourage the model to verify and refine its responses.
SVL is an ongoing research direction at Compiwer AI.
๐ Training
The final dataset contained 58,574 unique examples.
Dataset Examples
๐ฒ๐ฆ Moroccan Darija QA + Reasoning 2,170
๐ป OpenCodeReasoning-2 30,000
๐งฎ OpenMathInstruct-2 30,000
โก Model Specs
Property Value
Model Mtrini-SVL-1.1
Base Qwen3-VL-8B-Instruct
Parameters ~8B
Architecture Qwen3-VL
Training LoRA / PEFT
Precision BF16
Training Context 4096 tokens
Dataset 58,574 unique examples
Developer Compiwer AI
Origin ๐ฒ๐ฆ Salรฉ, Morocco
๐ Training Run
Steps: 3,661
Epochs: 1
Training Loss: 0.5855
Mean Token Accuracy: 83.55%
These metrics describe the training run and should not be treated as standardized benchmark scores.
๐ Usage
from transformers import Qwen3VLForConditionalGeneration, AutoTokenizer
model_id = "CompiwerAI/Mtrini-SVL-1.1-Merged"
tokenizer = AutoTokenizer.from_pretrained(
model_id,
trust_remote_code=True
)
model = Qwen3VLForConditionalGeneration.from_pretrained(
model_id,
torch_dtype="auto",
device_map="auto",
trust_remote_code=True
)
โ ๏ธ Limitations
Mtrini is an experimental open model.
It can hallucinate, generate incorrect code, misunderstand prompts, or produce incorrect mathematical answers.
Verify important outputs.
๐ Why Mtrini?
AI shouldn't only be built by the biggest companies on Earth.
Mtrini is Compiwer AI's experiment in building open AI with a focus on:
๐ฒ๐ฆ Moroccan language
๐ง Reasoning
๐ป Coding
๐งฎ Mathematics
๐ค Open AI
Building AI For Everyone.
Built in Salรฉ, Morocco ๐ฒ๐ฆ