deepspeed vs verl
Two of the top projects in Training and fine-tuning, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
deepspeed
PyTorch library for distributed training of large models with ZeRO
verl
Reinforcement learning training library for LLM post-training with PPO and GRPO
Verdict
deepspeed ranks higher, 79 to 68. Of the 3 score parts we measure for both, deepspeed leads on 1, verl on 1 and 1 is even.
- Adoptiondeepspeed leads by 32 points, 68 to 36: a wide gap.
- FreshnessEven at 100.
- MaintenanceWe do not score verl on this; deepspeed has 81.
- Agent-readyverl leads by 25 points, 70 to 45: a wide gap.
Pick deepspeed if you are
ML engineers training or fine-tuning large models on multi-GPU clusters
- Used to train models such as BLOOM 176B and Megatron-Turing NLG 530B
- Integrates with Hugging Face Transformers/Accelerate, Lightning, Composer, Determined and MMEngine
Pick verl if you are
ML engineers running RL post-training of LLMs on GPU clusters
- Pluggable backends: FSDP/FSDP2/Megatron-LM for training, vLLM/SGLang/HF for rollout
- Many RL algorithms built in: PPO, GRPO, ReMax, RLOO, REINFORCE++
| What we compare | deepspeed | verl |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 68, popular | 36, known |
| Freshness | 100, active | 100, active |
| Maintenance | 81, healthy | Not scored for this kind |
| Easy to run | Not scored for this kind | Not scored for this kind |
| Agent-ready | 45, minimal | 70, partly |
| Facts from GitHub and the README | ||
| Stars | 43.2k | 23.8k |
| License | Apache-2.0 (permissive) | Apache-2.0 (permissive) |
| Last commit | Oct 2026 | Oct 2026 |
| Last release | Sep 2026 | Sep 2026 |
| Language | Python | Python |
| Docker | Yes | Yes |
| GPU | Optional | Required |
| arm64 or Apple Silicon | Not stated | Not stated |
deepspeed
DeepSpeed is a PyTorch optimization library for training large deep learning models across many GPUs. It provides ZeRO memory partitioning with CPU and NVMe offload, 3D parallelism, Ulysses sequence parallelism and MoE support, configured through a JSON file. It integrates with Hugging Face Transformers and Accelerate, Lightning, MosaicML Composer, Determined and MMEngine.
Who it is for: ML engineers training or fine-tuning large models on multi-GPU clusters
Strengths
- Used to train models such as BLOOM 176B and Megatron-Turing NLG 530B
- Integrates with Hugging Face Transformers/Accelerate, Lightning, Composer, Determined and MMEngine
- Supports NVIDIA, AMD, Intel Gaudi/XPU/CPU, Huawei Ascend NPU and Tecorigin accelerators
- Installs from PyPI; C++/CUDA ops are JIT-compiled or pre-built
Weaknesses
- Needs PyTorch installed first and a CUDA or ROCm compiler for JIT ops
- Mainly tested on NVIDIA Pascal to Hopper and AMD MI100/MI200
- Windows lacks async IO and GDS support
- Large JSON config surface; ZeRO and offload tuning requires reading the docs
- GPU optional
- Docker
- Needs PyTorch, CUDA or ROCm compiler (nvcc/hipcc), ninja
- Models: PyTorch models, Hugging Face Transformers
verl
verl is the open-source implementation of the HybridFlow paper from ByteDance Seed. It runs RL and supervised fine-tuning for LLMs using FSDP, FSDP2 or Megatron-LM for training and vLLM, SGLang or HF Transformers for rollout generation. It covers PPO, GRPO, RLOO, ReMax and others, with multi-turn tool calling, VLM support and scaling to 671B models.
Who it is for: ML engineers running RL post-training of LLMs on GPU clusters
Strengths
- Pluggable backends: FSDP/FSDP2/Megatron-LM for training, vLLM/SGLang/HF for rollout
- Many RL algorithms built in: PPO, GRPO, ReMax, RLOO, REINFORCE++
- Scales to 671B models and hundreds of GPUs via expert parallelism
- Supports NVIDIA, AMD and Ascend hardware; LoRA RL to save memory
Weaknesses
- Needs multi-GPU clusters in practice; minimum hardware not stated in README
- Many algorithm recipes live in a separate verl-recipe repository and submodule
- Version-sensitive: README warns against vLLM 0.7.x and requires >=0.8.2
- Several features remain under verl/experimental until merged into the main library
- GPU required
- Docker
- Needs Ray, PyTorch, vLLM, SGLang, Megatron-LM
- Models: Qwen3.5, Qwen3, Qwen2.5, Qwen2.5-VL, Kimi-VL
More in Training and fine-tuning
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.