GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
valendra
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
valendra
Model Tree
openai/gpt-oss-120b
Input Modalities
Output Modalities
Supported Functionality
A short summary of the technique used to produce the checkpoint.
Expert pruning is the process of removing a fraction of the experts from each Mixture-of-Experts (MoE) layer of a language model while keeping the rest of the architecture intact. The router, the attention blocks, the embeddings, the tokenizer, and the chat template are unchanged. Only the per-layer expert table is reduced, which directly shrinks the model's parameter count, memory footprint and the amount of compute spent on the MoE block at inference time.
In this release the router weight matrix, the router bias and the
expert tensors were sliced along the expert dimension using
random-moe-pruning.
The default configuration was used (--prune-level 0.5, --top-k 4,
--num-examples 1000, --random-seed 42).
| Field | Value |
|---|---|
| Base model | openai/gpt-oss-120b |
| Original experts / layer | 128 |
| Pruned experts / layer | 64 |
| Experts kept per token | 4 |
| Prune level | 0.5 |
The metadata fields pruned_from_num_experts and
pruned_to_num_experts in config.json reflect the change for
traceability.
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizer model_id = "valendra/experimental-gpt-oss-120b-mini" tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained( model_id, dtype=torch.bfloat16, device_map="auto",)José Carlos García Ortega and Marta Barea Sepúlveda, PhD, Valendra Tech SL.
chat_template.jinja from the base model is reused.