# InternLM/lmdeploy

LMDeploy is a toolkit for compressing, deploying, and serving LLMs.

Repository: https://github.com/InternLM/lmdeploy
Canonical: https://ross.abutalabs.com/products/lmdeploy
Homepage: https://lmdeploy.readthedocs.io/en/latest
Language: Python
License: Apache-2.0
License Family: permissive
Topics: cuda-kernels, deepspeed, fastertransformer, llm-inference, turbomind, internlm, llama, llm, codellama, llama2, llama3
Last push: 2026-08-26T12:01:03+00:00

## Health v2 (maintenance only)
Score: 95/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 98, longevity 83
- inputs: {"age_days": 1175, "days_push": 7, "days_rel": 14, "gap_med": 20.0, "n_releases_24m": 33}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 8024, forks 728 (observed 2026-08-28T04:10:11.863585+00:00)

## What it is
LMDeploy is a toolkit for compressing, quantizing, deploying, and serving large language models, built around its high-performance TurboMind inference engine. It provides OpenAI-compatible serving APIs, CLI tools, and optimized CUDA kernels for fast LLM inference on NVIDIA GPUs.

## Use cases
- serve llm with openai compatible api
- quantize llama model to 4bit
- run deepseek v3 inference on gpu
- deploy chatbot backend for large language model
- speed up llm inference with turbomind
- compress and serve internlm models
- benchmark llm inference throughput

## When to choose
- you need high-throughput, low-latency LLM serving on NVIDIA GPUs
- you want built-in quantization (4bit, FP8, MXFP4) with a serving engine
- you deploy InternLM, Llama, Qwen, or DeepSeek models with an OpenAI-compatible API

## When to avoid
- you need CPU-only or non-NVIDIA hardware inference
- you only want to fine-tune or train models rather than serve them
- you prefer a simpler pure-PyTorch stack without custom CUDA kernels

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, gpu-computing, http-server, api-framework, chatbot
- domain: large-language-models, deep-learning, machine-learning, artificial-intelligence, gpu-computing, developer-tools
- platform: python, cli
- tags: turbomind, quantization, model-serving, cuda-kernels, openai-compatible-api, kv-cache, inference-engine, linux, gpu, docker

## Member repositories
- InternLM/lmdeploy (main) score 95

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:11.863585+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:31:49.674397+00:00, confidence not recorded.
  - readme: https://github.com/InternLM/lmdeploy (fetched 2026-08-28T04:10:11.863585+00:00, sha 57d0987acbf1)
- Data as of 2026-08-30T08:39:29.467469+00:00.