What if you could type a sentence in plain English —
and instantly get back the exact Excel formula you need?
No more searching StackOverflow.
No more struggling with nested parentheses.
No more wondering whether it’s COUNTIF, SUMIFS, or XLOOKUP this time.
That’s the idea behind 6SigmaMind, a lightweight AI assistant built on a tiny 1.7B parameter SLM (Small Language Model)… yet capable of generating real Excel formulas in seconds.
And yes — you can test it live right now.
👉 Try it here: https://huggingface.co/spaces/benkemp/6SigmaMindv2
🚀 Why 6SigmaMind Exists (and Why It’s Fun to Play With)
Most people think you need a massive model like GPT-5 to do anything useful.
But the AI world is changing fast: small models are getting smart enough to handle everyday tasks — insanely fast and completely free.
6SigmaMind is part experiment, part challenge:
How much Excel knowledge can we pack into a model tiny enough to run inside a Hugging Face Space?
Turns out: a surprising amount.
6SigmaMind can already write formulas for:
- SUMIFS
- COUNTIFS
- XLOOKUP
- IF + AND / OR
- Date math
- Text extraction
- Basic statistics (mean, stdev, correlation)
And because the model is so small, you get instant responses, even on CPU.
🔥 Try These Prompts Right Now
Copy one of these into the demo and see what happens:
1. Basic formulas
- “Write an Excel formula that sums numbers in column C.”
- “Count how many cells in column B contain the word ‘Closed’.”
2. Logic formulas
- “IF A2 is greater than 100 return ‘High’ otherwise return ‘OK’.”
- “Check if B2 is empty.”
3. Lookups
- “Lookup the price in column D where SKU in column A matches H2.”
- “Get the last non-empty value from column B.”
4. Statistics
- “Calculate the standard deviation of values in B2:B80.”
- “Return the correlation between columns A and B.”
You’ll immediately see where small models shine —
and where they need a little help.
That’s the fun part.
🧪 What Makes This Project Different?
6SigmaMind isn’t trying to replace ChatGPT.
It’s designed to answer one question:
👉 How far can a small model go if you focus it on one job: Excel formulas?
That means:
- faster responses
- lower compute
- more privacy
- experiment-friendly
- easy to fine-tune
- easy to embed in your own apps
And because it’s running on a public Hugging Face Space, anyone can test it in seconds, no login required.
🎯 Who Should Try 6SigmaMind?
You’ll enjoy this demo if you are:
- A data analyst
- A student
- A spreadsheet hobbyist
- A business manager
- A developer exploring SLMs
- Someone who hates writing long Excel formulas (we’ve all been there)
Even if you’re just curious about what tiny models can do —
you’ll likely be surprised.
6SigmaMind is powered by:
- HuggingFaceTB/SmolLM2-1.7B
- Custom prompt engineering
- A lightweight Gradio interface
- CPU-only inference (no heavy hardware needed)
No huge GPU.
No expensive API.
Just a tiny, efficient model doing a very specific job.
👇 One Ask: Break It. Push It. Confuse It.
Small models learn by being challenged.
So if you try the demo, please test it with:
- tricky nested formulas
- weird XLOOKUP conditions
- date calculations
- statistical formulas
- off-beat Excel tasks
Then share the results — good or bad.
It helps shape what 6SigmaMind becomes next.
👉 Try it now and see what a 1.7B Excel assistant can do:
https://huggingface.co/spaces/benkemp/6SigmaMindv2
🌟 What’s Coming Next?
This is only version 1.
Here’s what’s planned:
- “Formula-Only” strict mode
- Statistical reasoning expansion
- Multi-step tasks
- Model comparison dashboards
- Benchmarks vs Qwen, Phi-3, Gemma, etc.
- A fine-tuned version of 6SigmaMind trained on thousands of Excel interactions
Yes — we’re just getting started.
Nano Language Models 