The new J.O.S.I.E. version 2 family of personality-driven language models trained entirely on Apple Silicon
Gökdeniz Gülmez PRO
Goekdeniz-Guelmez
AI & ML interests
Transformers / NLP / Multimodal / Realtime M2M / J.O.S.I.E.-Home
Recent Activity
liked a dataset about 20 hours ago
lordx64/reasoning-distill-opus-4-7-max-sft liked a model 1 day ago
owensong/Inflect-Micro-v2 liked a model 3 days ago
SC117/LFM2.5-2.6B-Uncensored-GGUFOrganizations
Qwen3.5 but Gabliterated and Josiefied
The JOSIE models are a family of uncensored, high-performance language models developed by Gökdeniz Gülmez.
-
Gabliteration: Adaptive Multi-Directional Neural Weight Modification for Selective Behavioral Alteration in Large Language Models
Paper • 2512.18901 • Published • 6 -
Goekdeniz-Guelmez/JOSIE-4B-Instruct
Text Generation • 4B • Updated • 34 • • 8 -
Goekdeniz-Guelmez/JOSIE-4B-Thinking
Text Generation • 4B • Updated • 12 • • 14 -
mlx-community/JOSIE-4B-Instruct-bfloat16
Text Generation • 4B • Updated • 9
The next version of Abliteration
-
Gabliteration: Adaptive Multi-Directional Neural Weight Modification for Selective Behavioral Alteration in Large Language Models
Paper • 2512.18901 • Published • 6 -
Goekdeniz-Guelmez/Qwen3-4B-Instruct-2507-gabliterated
4B • Updated • 14 • 10 -
Goekdeniz-Guelmez/Qwen3-4B-Thinking-2507-gabliterated
4B • Updated • 42 • 9 -
Goekdeniz-Guelmez/Qwen3-4B-Sky-High-Hermes-gabliterated
4B • Updated • 12 • 13
Abliterated, and further fine-tuned to be the most uncensored models available.
-
Goekdeniz-Guelmez/Josiefied-Qwen3-30B-A3B-abliterated-v2
Text Generation • 31B • Updated • 69 • 23 -
Goekdeniz-Guelmez/Josiefied-Qwen3-14B-abliterated-v3
Text Generation • 15B • Updated • 90 • • 27 -
Goekdeniz-Guelmez/Josiefied-DeepSeek-R1-0528-Qwen3-8B-abliterated-v1
Text Generation • 8B • Updated • 23 • • 34 -
Goekdeniz-Guelmez/Josiefied-Health-Qwen3-8B-abliterated-v1
Text Generation • 8B • Updated • 9 • 12
Trained on opensourced and private custom DPO/ORPO datasets
A preview model of the new J.O.S.I.E. version 2 family of personality-driven language models trained entirely on Apple Silicon
The JOSIE-1.1 are the next iteration model family of uncensored, high-performance language models developed by Gökdeniz Gülmez.
JOSIE models using a custom dynamic Mixture of Expert architecture.
Other models Josiefied and Abliterated
-
Goekdeniz-Guelmez/Josiefied-Hermes-3-Llama-3.2-3B-v1
Text Generation • 3B • Updated • 15 • • 1 -
Goekdeniz-Guelmez/Josiefied-granite-4.0-micro-abliterated-v1
Text Generation • 3B • Updated • 14 • 3 -
Goekdeniz-Guelmez/Josiefied-Olmo-3-7B-Instruct-abliterated-v1
Text Generation • 7B • Updated • 10 • 2 -
Goekdeniz-Guelmez/Josiefied-Huihui-MoE-12B-A4B
Text Generation • 12B • Updated • 11
My (Gökdeniz Gülmez) first reasoning model fine-tuned on a custom distill dataset
Abliterated, and further fine-tuned to be the most uncensored models available.
-
Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-14B-Instruct-abliterated-v1
Text Generation • 15B • Updated • 351 • 3 -
Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1
Text Generation • 8B • Updated • 25 • 3 -
Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1-gguf
Text Generation • 8B • Updated • 166 • 1 -
Goekdeniz-Guelmez/Josiefied-Qwen2.5-14B-Instruct-abliterated-v4
Text Generation • 15B • Updated • 41 • • 20
-
Goekdeniz-Guelmez/J.O.S.I.E.v4o
Any-to-Any • Updated • 29 -
Goekdeniz-Guelmez/J.O.S.I.E.v4o-8b-stage1-beta2.2
Text Generation • 8B • Updated • 11 • 1 -
Goekdeniz-Guelmez/J.O.S.I.E.v4o-8b-stage1-beta1-Q4_k_m-gguf
8B • Updated • 14 • 1 -
Goekdeniz-Guelmez/J.O.S.I.E.v4o-8b-stage1-beta1-Q4_k_s-gguf
8B • Updated • 9 • 1
The new J.O.S.I.E. version 2 family of personality-driven language models trained entirely on Apple Silicon
A preview model of the new J.O.S.I.E. version 2 family of personality-driven language models trained entirely on Apple Silicon
Qwen3.5 but Gabliterated and Josiefied
The JOSIE-1.1 are the next iteration model family of uncensored, high-performance language models developed by Gökdeniz Gülmez.
The JOSIE models are a family of uncensored, high-performance language models developed by Gökdeniz Gülmez.
-
Gabliteration: Adaptive Multi-Directional Neural Weight Modification for Selective Behavioral Alteration in Large Language Models
Paper • 2512.18901 • Published • 6 -
Goekdeniz-Guelmez/JOSIE-4B-Instruct
Text Generation • 4B • Updated • 34 • • 8 -
Goekdeniz-Guelmez/JOSIE-4B-Thinking
Text Generation • 4B • Updated • 12 • • 14 -
mlx-community/JOSIE-4B-Instruct-bfloat16
Text Generation • 4B • Updated • 9
JOSIE models using a custom dynamic Mixture of Expert architecture.
The next version of Abliteration
-
Gabliteration: Adaptive Multi-Directional Neural Weight Modification for Selective Behavioral Alteration in Large Language Models
Paper • 2512.18901 • Published • 6 -
Goekdeniz-Guelmez/Qwen3-4B-Instruct-2507-gabliterated
4B • Updated • 14 • 10 -
Goekdeniz-Guelmez/Qwen3-4B-Thinking-2507-gabliterated
4B • Updated • 42 • 9 -
Goekdeniz-Guelmez/Qwen3-4B-Sky-High-Hermes-gabliterated
4B • Updated • 12 • 13
Other models Josiefied and Abliterated
-
Goekdeniz-Guelmez/Josiefied-Hermes-3-Llama-3.2-3B-v1
Text Generation • 3B • Updated • 15 • • 1 -
Goekdeniz-Guelmez/Josiefied-granite-4.0-micro-abliterated-v1
Text Generation • 3B • Updated • 14 • 3 -
Goekdeniz-Guelmez/Josiefied-Olmo-3-7B-Instruct-abliterated-v1
Text Generation • 7B • Updated • 10 • 2 -
Goekdeniz-Guelmez/Josiefied-Huihui-MoE-12B-A4B
Text Generation • 12B • Updated • 11
Abliterated, and further fine-tuned to be the most uncensored models available.
-
Goekdeniz-Guelmez/Josiefied-Qwen3-30B-A3B-abliterated-v2
Text Generation • 31B • Updated • 69 • 23 -
Goekdeniz-Guelmez/Josiefied-Qwen3-14B-abliterated-v3
Text Generation • 15B • Updated • 90 • • 27 -
Goekdeniz-Guelmez/Josiefied-DeepSeek-R1-0528-Qwen3-8B-abliterated-v1
Text Generation • 8B • Updated • 23 • • 34 -
Goekdeniz-Guelmez/Josiefied-Health-Qwen3-8B-abliterated-v1
Text Generation • 8B • Updated • 9 • 12
My (Gökdeniz Gülmez) first reasoning model fine-tuned on a custom distill dataset
Abliterated, and further fine-tuned to be the most uncensored models available.
-
Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-14B-Instruct-abliterated-v1
Text Generation • 15B • Updated • 351 • 3 -
Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1
Text Generation • 8B • Updated • 25 • 3 -
Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1-gguf
Text Generation • 8B • Updated • 166 • 1 -
Goekdeniz-Guelmez/Josiefied-Qwen2.5-14B-Instruct-abliterated-v4
Text Generation • 15B • Updated • 41 • • 20
Trained on opensourced and private custom DPO/ORPO datasets
-
Goekdeniz-Guelmez/J.O.S.I.E.v4o
Any-to-Any • Updated • 29 -
Goekdeniz-Guelmez/J.O.S.I.E.v4o-8b-stage1-beta2.2
Text Generation • 8B • Updated • 11 • 1 -
Goekdeniz-Guelmez/J.O.S.I.E.v4o-8b-stage1-beta1-Q4_k_m-gguf
8B • Updated • 14 • 1 -
Goekdeniz-Guelmez/J.O.S.I.E.v4o-8b-stage1-beta1-Q4_k_s-gguf
8B • Updated • 9 • 1