Bronze
Back the work
$10 / month
- Your name on our supporters list
- A vote on what we build next
- A thank you in the build log
Open-weight model
II-Medical 8B
An open AI model for medical questions and clinical text. It runs fully offline on a small desktop, so patient details never leave the clinic.
- Field
- Medicine
- Runs
- Fully offline
- Built on
- II-Medical 8B
- License
- Apache-2.0, free
II-Medical 8B
II-Medical 8B is an open AI model for medical questions and clinical text. It runs fully offline on a small desktop, so patient details never leave the clinic.
What it can do
It answers health and medical questions, explains conditions and terms in plain words, and works through clinical text step by step. It is built on Intelligent-Internet’s II-Medical-8B, which learned to reason its way to an answer rather than just guess, and it is packed into ready-to-run files for a single desktop.
How well it works
We scored five builds on MedMCQA, a 50-question medical exam test, on a small Spark desktop. The Q5_K_M build scored the best at 52 percent, above the full-size build, while running at about 36 tokens a second. The table above shows every build.
This is a short test, and the model is a study and drafting helper, not a doctor. It can be wrong, so never use it to make a real medical decision. Always check with a qualified clinician.
How to run it
Download the GGUF files (the ready-to-run format) and run them with llama.cpp on a Spark-class desktop, a small AI machine with 128 GB of memory. The Q5_K_M build is the sweet spot here: the best score and a healthy 36 tokens a second.
Install
huggingface-cli download Orionfold/II-Medical-8B-GGUFllama-cli -hf Orionfold/II-Medical-8B-GGUF:Q5_K_MUse it
llama-cli -hf Orionfold/II-Medical-8B-GGUF:Q5_K_M -p "Explain the difference between Type 1 and Type 2 diabetes."from llama_cpp import Llama
llm = Llama.from_pretrained(
repo_id="Orionfold/II-Medical-8B-GGUF",
filename="*Q5_K_M.gguf",
)
out = llm("Explain the difference between Type 1 and Type 2 diabetes.")
print(out["choices"][0]["text"])
Specs
- Base model
- Intelligent-Internet/II-Medical-8B
- Format
- GGUF (ready to run)
- Builds
- Q4_K_M · Q5_K_M · Q6_K · Q8_0 · F16
- Best build
- Q5_K_M (about 36 tokens a second on a Spark desktop)
- License
- Apache-2.0 (free to use)
Benchmarks
| Build | MedMCQA score | Speed on a Spark |
|---|---|---|
| Q4_K_M | 42% | 44 tokens a second |
| Q5_K_M (best pick) | 52% | 36 tokens a second |
| Q6_K | 46% | 33 tokens a second |
| Q8_0 | 48% | 28 tokens a second |
| F16 (full size) | 48% | 16 tokens a second |
Used in the open
Live counts from HuggingFace, refreshed when the site builds. Built and maintained in the open by Orionfold.
- 455
- Downloads · last 30 days
Sponsor II-Medical 8B
Back this work with a monthly tier. Your support moves your requests up the list, and Gold or Platinum earns a badge on the roadmap item you back.
Silver
Get a say
$25 / month
- Everything in Bronze, plus:
- Your requests join the priority line
- Early access to new versions
Gold
Move it up the list
$50 / month
- Everything in Silver, plus:
- High priority on your feature and fix requests
- A Gold badge on the roadmap item you back
- A monthly call to plan the work
Platinum
Shape the roadmap
$100 / month
- Everything in Gold, plus:
- Top of the line, we work your requests first
- A Platinum badge on the roadmap item you back
- A direct line to us and a hand in the roadmap
Need something specific? Send an enquiry from the roadmap.
Keep exploring
Model
Patent Strategist
Offline patent reasoning in ready-to-run files, built with the NeMo toolkit. Nothing leaves your desktop.
Book
AI Research on NVIDIA DGX Spark
Real notes from doing AI research on one desktop. The NVIDIA DGX Spark is a small machine with huge power (petascale means it runs about a quadrillion math steps a second), so you can push local AI further with no cloud needed. Every lesson is backed by code that runs.