Vinci MLE 1.0 · Research model · Open weights
MLE 8B 1.0
A smaller research model trained to help investigate problems while building AI. Developed in Canada by SimpleDirect®, using IBM Granite 4.1 8B as its foundation.
What it is trained to help with
It is trained to read files and experiment results, look for problems, and explain what to do next.
The training uses 91 examples of multi-step work across 64 tasks. It covers eight areas: data quality; understanding experiments; restarting failed training; managing memory and batches; checking how results are scored; unstable training; packaging and running models; and matching model settings.
Change it. Leave it. Ask for more.
- Make a change
- Suggest a limited fix when the evidence supports it.
- Leave it alone
- Keep a working setup as it is when a change is not justified.
- Ask for more evidence
- Explain what is missing instead of guessing at a repair.
This describes the training, not proven results on new tasks.
What we have—and haven’t—tested
We’ve tested coding and other general skills. We haven’t yet completed a valid test on new AI-engineering tasks. We plan to require those tests before releasing MLE 1.1.
The model cards show coding, tool-use, and reasoning results compared with the original Granite models. Those tests do not tell us how well MLE solves new AI-engineering problems.
In one fixed 8B smoke-test case, the model produced an invalid tool name. Validate tool names and distinguish tool calls from final answers; the model card documents the case.
See the full results and limitationsRun it on your own setup
The model files, setup examples, and technical guides are on Hugging Face. You can choose the full weights or smaller GGUF versions. The download pages explain the available formats and their checks.
These are downloadable research models, not a ready-made app or hosted assistant. To let a model work with files or run commands, your application needs to provide the tools, a controlled workspace, and permissions. Review and test any changes it suggests.
Choose 8B for a smaller download and less memory use. 30B needs more memory and scored higher in our coding and tool-use tests. We don’t yet know which is better at AI-engineering tasks.