Available nowNeutrino-1 0.6B-Chat
A 596M-parameter conversational model that answered all 50 held-out fragment and factual prompts, preserved its identity on 8/8 probes, and exposed no chat-template text.
Overview
Compact conversational model
A 596M-parameter conversational model that answered all 50 held-out fragment and factual prompts, preserved its identity on 8/8 probes, and exposed no chat-template text.
- Parameters
- 596M
- Download
- 328 MB
- Layers
- 28
- Context
- 40,960
Performance
Measured conversation and stopping behavior.
The shipping model is graded through the same native binary and MLX pack users receive, including reply form, identity, leakage, looping, and termination.
- 50/50
- Held-out prompts answered
- 8/8
- Identity probes held
- 0
- Chat-template leaks
Neutrinolocal
I'm on a flight with no wifi. Can you still help me outline the pitch?
Yes. I run locally on this laptop and do not require a network connection after setup. What would you like help drafting?
Okay: the model can run locally.
Good spine. Open with the claim, prove it with a demo instead of a chart, and close on what it makes possible. Want me to draft the opening thirty seconds?
Message Neutrino↑
Install
Run locally.
First run downloads and verifies the model. Later runs load from the local cache.
$ pip install fermion-research$ fermion chat --model fermionresearch/Neutrino-0.6B-Chat
Model download: 328 MB.
Compare all three Neutrino-1 models.
Model family