Open Source Alternatives
Open source alternatives to DeepSpeed
Open source alternatives to DeepSpeed, ranked by GitHub stars and freshness.
12 open-source alternatives in the index, ranked by GitHub stars and freshness.
Colossal-AI
Community
Making large AI models cheaper, faster and more accessible
Alternative to Deepspeed, Megatron Lm
Best for: Teams training large models who have access to multiple GPUs and need to optimize resource efficiency
NeMo Framework
Community
A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech
Alternative to Megatron Lm, Deepspeed
Best for: Researchers and ML engineers building custom LLMs, speech systems, or multimodal models who need low-level control and scalability.
Megatron-LM
Community
Ongoing research training transformer models at scale
Alternative to Deepspeed, Colossal Ai
Best for: ML engineers training large transformer models who need production-grade distributed training infrastructure
Horovod
Community
Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.
Alternative to Deepspeed
Best for: ML engineers training large models who need to scale across multiple GPUs or nodes without rewriting training logic
Liger-Kernel
Community
Efficient Triton Kernels for LLM Training
Alternative to Deepspeed, Colossal Ai, Unslothai
Best for: Developers training large language models who want to optimize throughput and memory without rewriting their training loops.
torchtitan
Community
A PyTorch native platform for training generative AI models
Alternative to Deepspeed, Megatron Lm
Best for: Teams using PyTorch to train custom generative AI models at scale
nanotron
Community
Minimalistic large language model 3D-parallelism training
Alternative to Deepspeed, Colossal Ai, Megatron Lm
Best for: Researchers and engineers who need a simple, hackable framework for distributed LLM training experiments.
maxtext
Community
A simple, performant and scalable Jax LLM!
Alternative to Megatron Lm, Deepspeed, Colossal Ai
Best for: Developers already using Jax who need a streamlined, scalable LLM training framework
Mesh Tensorflow
Community
Mesh TensorFlow: Model Parallelism Made Easier
Alternative to Deepspeed, Megatron Lm, Colossal Ai
Best for: Developers training large transformer models on TPU or GPU clusters
MInference
Community
[NeurIPS'24 Spotlight, ICLR'25, ICML'25] To speed up Long-context LLMs' inference, approximate and dynamic sparse calculate the attention, which reduces inference latency by up to
Alternative to Vllm, Sglang, Tensorrt Llm +1 more
Best for: Developers optimizing long-context LLM inference on NVIDIA GPUs
BMTrain
Community
Efficient Training (including pre-training and fine-tuning) for Big Models
Alternative to Deepspeed, Colossal Ai, Megatron Lm
Best for: Developers and researchers training or fine-tuning large models who need a specialized, efficiency-focused Python framework
OneComp
Community
Python package for LLM compression
Alternative to Deepspeed, Colossal Ai
Best for: Developers needing to shrink LLMs for deployment on limited hardware