Using our LEM Runtime, a model that fits on your phone, wrote this over 10 turns, no history reply, just the next turn.
Model: huggingface.co/lthn/lemer a Gemma4-E2B-IT retune.
New revision, our first was using day-1 forks.
huggingface.co/lthn/lemer
ollama run hf.co/lthn/lemer:Q4_K_M
Apple Native: mlx_lm.server --model "lthn/lemer"
LM Studio, just click below with it open.
We want to introduce, Lemer.
The first of 4 LetheanEthicalModel (LEM) models for Gemma4.
Adapters to re-introduce the ethics speciality in our Gemma3 models to come later.
Compatable with EU AI laws, EUPL-1.2, requires around 4GB mem 9.5 for bf16.
huggingface.co/lthn/lemer
We want to introduce, Lemer.
The first of 4 LetheanEthicalModel (LEM) models for Gemma4.
Adapters to re-introduce the ethics speciality in our Gemma3 models to come later.
Compatable with EU AI laws, EUPL-1.2, requires around 4GB mem 9.5 for bf16.
huggingface.co/lthn/lemer
To little fan fair today, we would like to introduce.
Lethean Ethical Modal
huggingface.co/datasets/lthn…
The LEK Method: Ethical Kernel Fine-Tuning as an Alternative to RLHF Behavioural Conditioning: github.com/LetheanNetwork/LE…
Also, evendece Unsenored is dangeours & drops quality.
Welcome to a better Google search; literally.
Why not try an AI from the EU, guarded by strict privacy laws?
Your questions are not used for training, period, end of.
(not even an affiliate-linked promo, it's just good, I think you should try it)
lumo.proton.me/about