Skip to content

Instantly share code, notes, and snippets.

@iamdejan
Created March 26, 2026 10:03
Show Gist options
  • Select an option

  • Save iamdejan/66289ad82c716d193f511bb813789b65 to your computer and use it in GitHub Desktop.

Select an option

Save iamdejan/66289ad82c716d193f511bb813789b65 to your computer and use it in GitHub Desktop.
Model parameters for math testings.
Model Context Length GPU Offload CPU Thread Pool Size Evaluation Batch Size Number of Experts Number of layers to force the experts to CPU Temperature Top K Sampling Repeat Penalty Top P Sampling Min P Sampling
allenai/Olmo-3-7B-Think 7729 32 8 512 - - 0.6 40 1.1 0.95 0.05
lm-provers/QED-Nano 84878 36 8 784 - - 0.6 40 1.1 0.95 0.05
mradermacher/Nemotron-Cascade-2-30B-A3B-GGUF 16384 52 8 784 8 8 1 40 1.1 0.95 0.05
microsoft/Phi-4-mini-reasoning 60319 32 8 784 - - 0.8 40 1.1 0.95 0.05
bartowski/nvidia_OpenMath-Nemotron-14B-GGUF 25844 48 8 784 - - 0.6 40 1.1 0.95 0.05
inclusionAI/Ring-mini-2.0 4096 20 8 784 8 0 0.6 40 1.1 0.95 0.05
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment