This repository contains the Q4_K_M GGUF artifact for GLM-4.7 Flash with native
MTP metadata and tensors.
For upstream architecture details, chat template guidance, sampling
recommendations, license terms, and benchmark notes, see the source model card:
zai-org/GLM-4.7-Flash.
1# Download the Q4_K_M MTP GGUF artifact.
2hf download meshllm/GLM-4.7-Flash-MTP-GGUF \
3 --include 'GLM-4.7-Flash-MTP-Q4_K_M.gguf'
1# Run locally with Mesh LLM.
2mesh-llm serve --model "meshllm/GLM-4.7-Flash-MTP-GGUF"
1# Check the local OpenAI-compatible model list.
2curl -s http://localhost:3131/v1/models
1# Send an OpenAI-compatible chat request.
2curl -s http://localhost:3131/v1/chat/completions \
3 -H "Content-Type: application/json" \
4 -d '{
5 "model": "meshllm/GLM-4.7-Flash-MTP-GGUF",
6 "messages": [{"role": "user", "content": "Write a tiny hello-world function in Rust."}],
7 "max_tokens": 128
8 }'
MTP, or multi-token prediction, lets a model predict the next token and an
extra future token in the same decode step. That gives the runtime a candidate
sequence to verify instead of waiting for one full model pass per token.
