GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
nightmedia
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
nightmedia
Model Tree
google/gemma-4-12B-it
Input Modalities
Output Modalities
Supported Functionality
This is a Nightmedia release, trained with locally generated content.
This training includes TNG-infused coding lessons.
The teacher model used was the qx86-hi quant of the nightmedia/Qwen3.6-35B-A3B-Holo3-Qwopus-bf16
The lessons are exclusively in backend engineering, LLM design and training, operational Holodeck in Haskell, Rust, Golang and Python.
The set up was in the DS9 Holodeck, with in-person commentary by Spock, Data, Quark, Worf, Odo, Julian, Garak, Q, and many others.
Additionally, some of the Polaris Alpha questions were used to distill from the teacher model.
brainwaves
arc arc/e boolq hswag obkqa piqa winobf16 0.553,0.771,0.849mxfp8 0.551,0.779,0.865q8-hi 0.552,0.770,0.848qx86-hi 0.539,0.771,0.835mxfp4 0.505,0.750,0.853Quant Perplexity Peak Memory Tokens/secbf16 67.645 ± 1.080 30.95 GB 549mxfp8 74.437 ± 1.177 19.42 GB 373q8-hi 65.357 ± 1.031 20.54 GB 460qx86-hi 64.724 ± 1.021 19.00 GB 394mxfp4 117.640 ± 2.029 13.47 GB 442
gemma-4-12B-TNG-V6-q8-hi-mlx
brainwaves
arc arc/e boolq hswag obkqa piqa winoq8-hi 0.524,0.713,0.775Quant Perplexity Peak Memory Tokens/secq8-hi 75.792 ± 1.251 20.54 GB 446
gemma-4-12B-TNG-V1-q8-hi-mlx
brainwaves
arc arc/e boolq hswag obkqa piqa winomxfp8 0.439,0.617,0.695q8-hi 0.475,0.615,0.724qx86-hi 0.468,0.614,0.753Quant Perplexity Peak Memory Tokens/secmxfp8 85.212 ± 1.319 19.42 GB 360q8-hi 73.981 ± 1.145 20.54 GB 373qx86-hi 71.764 ± 1.103 19.00 GB 416
gemma-4-12B-it
brainwaves
arc arc/e boolq hswag obkqa piqa winomxfp8 0.385,0.527,0.766,0.509,0.386,0.664,0.579q8-hi 0.394,0.522,0.774,0.520,0.370,0.674,0.583qx86-hi 0.391,0.526,0.777,0.521,0.366,0.677,0.590mxfp4 0.371,0.517,0.660,0.508,0.368,0.677,0.581Quant Perplexity Peak Memory Tokens/secmxfp8 175.766 ± 3.092 19.42 GB 463mxfp4 283.801 ± 5.361 13.47 GB 484
The V8 iteration introduces a random pick of Claude code traces from angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k-raw mixed in with a random pick of lessons from the first stage.
-G