Model Surgeon
You use neural networks every day. You've never actually seen one.
Open one and rotate its real anatomy, layer by layer, in your browser. It's the difference between believing things about AI and knowing them.
- Parameters
- 67,556,476
- Layers
- 78
- Tensors
- 109
- Size
- 270.2 MB
Delete a piece of a live model. Watch it keep talking.
You stop guessing what a layer does and start knowing what it cost. Every cut can be undone, and every consequence is measured before you commit.
Excise
Remove a layer and see exactly what it cost in size, speed and behaviour.
Graft
Swap in a better part without rebuilding the model around it.
Chimera Ward
Blend two models into one and decide how much of each survives.
Tissue imaging
Watch a real input light the network up, and spot the dead weight in red.
Hardware costing
Know what every edit saves on the hardware you actually deploy to.
Your next model ships with receipts, not “trust me.”
Every incision is written down in order and sealed into what you export. When someone asks what changed, you send a link instead of an explanation.
- loadmodel.safetensors · 270.2 MBdone
- forensicsno anomalies foundclean
- testforward pass on the edited graphpass
- imaging16 synthetic tokens → [1, 512]weak
- exportlineage.intoto.jsonl · ed25519signed
Anyone on earth can check it. Free, forever.
Flip a single byte and the verifier says tampered. It runs offline, needs no account, and costs nothing, for everyone, forever.
pip install falconsai-surgeonsurgeon verify package.zip- Parsed, never executed
- Models deleted when the session ends
- No code, no install
- From the team behind 1.3B model downloads
Before you open it.
What is Model Surgeon?
A visual operating theatre for neural networks in your browser: load a real model file, see its anatomy in 3-D, operate with reversible tools, measure every consequence, and export with a cryptographic record anyone can verify.
Do I need to write code?
No. Every operation is point-and-click. Code only appears in what you export — scripts and a notebook you can run elsewhere.
Is it safe to open a model from an unknown source?
Yes — that's a core design promise. Models are parsed like documents, never executed, so a booby-trapped checkpoint has nothing to run.
What happens to my model when I close the tab?
Everything is purged — model, edits, generated assets. Sessions are ephemeral by design. Use Save Theatre first if you want to resume later.
What does it cost?
Free with signup, no card: full surgery and measurement on models up to 1 GB. Practitioner is $39/month or $390/year and raises the ceiling to 20 GB, with signed audit-grade exports, live validation and private Hugging Face pushes. Verifying a package is free for everyone, forever.
How big a model can I load?
Free accounts open models up to 1 GB. Practitioner raises the ceiling to 20 GB for .safetensors files, which stream to disk and are read piece-by-piece. Other formats cap at 1 GB in every tier because they must be fully unpacked in memory.
Open
the theatre
Free to start. No card, no install. Verification is free forever.