GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Nickyang
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
Nickyang
Model Tree
zai-org/GLM-4.7-Flash
Input Modalities
Output Modalities
Supported Functionality
python
from transformers import AutoModelForCausalLM, AutoTokenizer model_id = "Nickyang/GLM-4.7-Flash-RAZOR-17B-A3B-E32of64"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained( model_id, dtype="bfloat16", device_map="auto",) messages = [{"role": "user", "content": "Explain mixture-of-experts routing."}]inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, return_tensors="pt",).to(model.device)print(tokenizer.decode(model.generate(inputs, max_new_tokens=256)[0]))Requires a Transformers build that contains the native glm4_moe_lite
implementation. The checkpoint stores bfloat16 weights in five safetensors
shards.
RAZOR asks whether the surviving computation can replace an expert's function, rather than how often or how strongly the expert fires. For a token routed to the set S with normalized weights wj, let c=∑j∈Swjfj be the routed mixture and rj=fj−c the consensus residual of expert j. Deleting a selected expert i makes the router promote its highest-ranked unselected expert r, whose pseudo-weight wr is its score divided by the original selected score sum. With λ the routed-output scale, the exact local output change is
δi=λ∥c−c~−i∥2=λScores are aggregated by conditional root mean square over the calibration tokens routed to each expert, and the highest-scoring experts are retained per layer. See the paper for the derivation and the code for the implementation.
Calibration used 32,768-token rows from RazorCal, the 2,048-sample multi-domain corpus released with RAZOR.
bash
pip install razor-moe razor saliency --model <path-to-GLM-4.7-Flash> \ --data data/RazorCal.json --max-len 32768 --out out/salrazor prune --model <path-to-GLM-4.7-Flash> --saliency out/sal \ --method razor --ratio 0.5 --out out/prunedkept_expert_indices.json in this repository is the keep-set manifest this
checkpoint was built from, in the format razor verify expects:
bash
razor verify --pruned . --src <path-to-GLM-4.7-Flash>Selection depends on the calibration draw, so an independent run reproduces the procedure rather than this exact expert set.
Expert pruning is lossy. Benchmark retention and predictive fidelity do not guarantee stable generation: the paper reports that responses change in diversity, formatting and termination behaviour even where task accuracy is largely preserved. Evaluate on your own workload before deploying.
The calibration corpus is multi-domain but finite, so behaviour on domains far from it is not characterised by the released measurements. This repository publishes the 50% expert-removal setting only; the paper also reports 25%.
This is a derivative of GLM-4.7-Flash, released by Z.ai under the MIT license, and is distributed under the same license. The retained weights are the base model's own weights; RAZOR contributes the expert selection, not new parameters.
The RAZOR code is Apache-2.0. RazorCal records remain subject to their applicable upstream terms; see data/LICENSE-DATA.
bibtex
@misc{song2026razorpruningreplaceableexperts, title={RAZOR: Pruning Replaceable Experts in LLMs}, author={Mingyang Song and Mao Zheng}, year={2026}, eprint={2609.30465}, archivePrefix={arXiv}, primaryClass={cs.LG}, url={https://arxiv.org/abs/2609.30465},}