voidx-decider-12b-GGUF

Q4_K_M GGUF of voidx-decider-12b: the LoRA adapter merged into google/gemma-4-12B-it (revision 707f0a3b), then quantized with llama.cpp. This is not an official Google model.

File Size sha256
voidx-decider-12b-Q4_K_M.gguf 7.38 GB c48d1fc0851d5bf368a93e07efe725dd2235145e5dbd8877df9eeeaf408195d1

Research release. The model did not pass its locked-test gate and is highly vulnerable to prompt injection. See the main model card for usage, evaluation and limitations.

  • Q4_K_M vs bf16 on the locked test: accuracy 78.4% vs 79.3%, ECE 0.034 vs 0.028.
  • Serving temperature: T = 0.98580.
  • llama-server: use /completion with n_probs, or pass chat_template_kwargs: {"enable_thinking": false} on the chat endpoint.

License: Apache-2.0. See LICENSE and NOTICE.

Downloads last month
25
GGUF
Model size
12B params
Architecture
gemma4
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for upfall-co/voidx-decider-12b-GGUF

Quantized
(1)
this model