Skip to content
xLLM
PR preview
Getting Started
Hardware
User Guide
CookBook
Developer Guide
CLI Reference
Search
Ctrl
K
Cancel
GitHub
EN
EN
中
Home
Getting Started
Quick Start
Launch xllm
Supported Models
Hardware
Hardware Platforms
Ascend NPU
Cambricon MLU
Hygon DCU
MetaX MACA
Mthreads MUSA
NVIDIA GPU
User Guide
Basic Usage
Multi-Machine Deployment
Online Service
OpenAI-Compatible APIs
OpenAI APIs - Vision
OpenAI APIs - Embeddings
Anthropic-Compatible API
Offline Inference
Advanced Features
Async schedule
Multi-stream parallel
ChunkedPrefill Scheduler
Zero Evict Scheduler
Disaggregated PD
Prefix Cache Optimization
Global Multi-Level KV Cache
Virtual Memory for KV Cache and Model Weights
EP Parallelism
MoE Load Balancing (EPLB)
MTP Speculative Inference
Graph Mode
FlashComm
Adaptive Speculation
xLLM Service
Overview
Build and run
KV events and CAR
CookBook
Autoregressive Models
Qwen
Qwen3.5
Qwen3
Qwen3-Next
Qwen3-VL
Qwen2.5-VL
DeepSeek
DeepSeek-V4
DeepSeek-V3.2
DeepSeek-V3.1
DeepSeek-V3
DeepSeek-R1
GLM
GLM-5.3-Flash
GLM-5.2
GLM-5.1
GLM-5
GLM-4.7
GLM-4.7-Flash
GLM-4.6
GLM-4.6V
GLM-4.5
GLM-4.5V
Kimi
Kimi2
Kimi-K2.5 / Kimi-K2.6
MinMax
MiniMax-M2.7
Diffusion Models
Flux
Flux
Flux2
Wan
Wan2.1
Wan2.2
Qwen-Image
Qwen-Image
Developer Guide
Code Architecture and Request Flow
C++ Serving and Python Model Execution
Graph Execution and Memory Reuse
Generative Recommendation Inference Architecture
AI-Assisted Development Workflow
xLLM-Kernel Development Guide
TileLang Kernel Development on Ascend NPU
Extending Auto Config
Accuracy Evaluation with EvalScope
Inference Benchmarking with EvalScope
Profiling a Running Service
CLI Reference
GitHub
Select language
EN
中
GLM-5.1
Copy page
GLM-5.1 inference recipes are consolidated into
GLM-5 / GLM-5.1 / GLM-5.2
.
Report issue
xLLM PR preview xLLM PR preview