--- title: "horovod vs Megatron-LM" type: "comparison" canonical_url: "https://www.graphcanon.com/compare/horovod-horovod-vs-nvidia-megatron-lm" tools: ["horovod-horovod", "nvidia-megatron-lm"] --- # horovod vs Megatron-LM *GraphCanon updated Jul 12, 2026* ## Verdict Pick horovod when tags unique to horovod: baidu, deep-learning, deeplearning, keras; pick Megatron-LM when requirements: Min 32 GB RAM; Requires NVIDIA GPUs for optimized performance. Non-GPU usage is not supported or recommended.; Installation from source can be resource-intensive and may require limiting parallel compilation jobs to avoid running out of memory.. [horovod](http://horovod.ai) reports 15k GitHub stars, 2.2k forks, and 406 open issues, last pushed Jun 20, 2026. [Megatron-LM](https://docs.nvidia.com/megatron-core/developer-guide/latest/get-started/quickstart.html) has 17k stars, 4.2k forks, and 988 open issues, last pushed Jul 11, 2026. Figures are from public GitHub metadata via [horovod's repository](https://github.com/horovod/horovod) and [Megatron-LM's repository](https://github.com/NVIDIA/Megatron-LM). | | [horovod](/tools/horovod-horovod.md) | [Megatron-LM](/tools/nvidia-megatron-lm.md) | | --- | --- | --- | | Tagline | Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet. | Ongoing research training transformer models at scale | | Stars | 14,692 | 17,020 | | Forks | 2,238 | 4,219 | | Open issues | 406 | 988 | | Language | Python | Python | | Adopt for | - | Megatron-LM from NVIDIA is a research-focused tool for developing and training large-scale language models with transformer architectures, emphasizing efficient parallelism across multiple GPUs. | | Persona | - | - | | Runtime | - | - | | License | Other | Other | | Categories | Model Training | Model Training | ## Trust and health _Sourced signals - not a safety guarantee. No winner column._ | | [horovod](/tools/horovod-horovod.md) | [Megatron-LM](/tools/nvidia-megatron-lm.md) | | --- | --- | --- | | Maintenance | Active (82%) | Very active (96%) | | Days since push | 21d | 0d | | Open issues (now) | 406 | 988 | | Full report | [trust report](/tools/horovod-horovod/trust.md) | [trust report](/tools/nvidia-megatron-lm/trust.md) | ## Decision facts: Megatron-LM - **Requirements:** Min 32 GB RAM; Requires NVIDIA GPUs for optimized performance. Non-GPU usage is not supported or recommended.; Installation from source can be resource-intensive and may require limiting parallel compilation jobs to avoid running out of memory. - **Adopt for:** Megatron-LM from NVIDIA is a research-focused tool for developing and training large-scale language models with transformer architectures, emphasizing efficient parallelism across multiple GPUs. ## Choose when ### Choose horovod if… - Tags unique to horovod: baidu, deep-learning, deeplearning, keras. - Leaner open-issue backlog (406). ### Choose Megatron-LM if… - Requirements: Min 32 GB RAM; Requires NVIDIA GPUs for optimized performance. Non-GPU usage is not supported or recommended.; Installation from source can be resource-intensive and may require limiting parallel compilation jobs to avoid running out of memory.. - Tags unique to Megatron-LM: large-language-models, model-para, transformers. - The tool is particularly beneficial when your project is GPU-centric and benefits from advanced parallelism techniques such as Tensor, Pipeline, Data, Expert, and Cluster Parallelisms (TP, PP, DP, EP, ## When NOT to use horovod - Model Training: Try prompting and RAG first; fine-tuning is the answer to style/format, not missing knowledge. ## When NOT to use Megatron-LM - Avoid Megatron-LM if your computational setup does not include NVIDIA GPUs as it leverages GPU-specific features and parallelisms that may not be available or efficient on non-NVIDIA hardware. - If you need portability across various hardware without depending on proprietary optimizations, other tools might better serve your needs. ## Common questions ### What is the difference between horovod and Megatron-LM? horovod: Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.. Megatron-LM: Ongoing research training transformer models at scale. See the comparison table for live GitHub stats and shared categories. ### When should I choose horovod over Megatron-LM? Choose horovod over Megatron-LM when Tags unique to horovod: baidu, deep-learning, deeplearning, keras; Leaner open-issue backlog (406). ### When should I choose Megatron-LM over horovod? Choose Megatron-LM over horovod when Requirements: Min 32 GB RAM; Requires NVIDIA GPUs for optimized performance. Non-GPU usage is not supported or recommended.; Installation from source can be resource-intensive and may require limiting parallel compilation jobs to avoid running out of memory.; Tags unique to Megatron-LM: large-language-models, model-para, transformers; The tool is particularly beneficial when your project is GPU-centric and benefits from advanced parallelism techniques such as Tensor, Pipeline, Data, Expert, and Cluster Parallelisms (TP, PP, DP, EP,. ### When should I avoid horovod? Model Training: Try prompting and RAG first; fine-tuning is the answer to style/format, not missing knowledge. ### When should I avoid Megatron-LM? Avoid Megatron-LM if your computational setup does not include NVIDIA GPUs as it leverages GPU-specific features and parallelisms that may not be available or efficient on non-NVIDIA hardware. If you need portability across various hardware without depending on proprietary optimizations, other tools might better serve your needs. ### Is horovod or Megatron-LM more popular on GitHub? Megatron-LM has more GitHub stars (17,020 vs 14,692). Stars measure visibility, not whether either tool fits your constraints. ### Are horovod and Megatron-LM open source? Yes - both are open-source projects on GitHub (horovod: Other, Megatron-LM: Other). ### Where can I find alternatives to horovod or Megatron-LM? GraphCanon lists graph-backed alternatives at [horovod alternatives](/tools/horovod-horovod/alternatives) and [Megatron-LM alternatives](/tools/nvidia-megatron-lm/alternatives) ([horovod markdown twin](/tools/horovod-horovod/alternatives.md), [Megatron-LM markdown twin](/tools/nvidia-megatron-lm/alternatives.md)), ranked by typed relationship edges rather than popularity votes. ### Is there a machine-readable version of this comparison? Yes. The markdown twin at [this comparison](/compare/horovod-horovod-vs-nvidia-megatron-lm.md) mirrors this page for agents and LLM crawlers, with the same stats table and FAQ answers. ### Which is better maintained, horovod or Megatron-LM? horovod: Active. Megatron-LM: Very active. Compare maintenance labels, days since push, and release cadence in the trust section below - stars alone do not measure maintenance. ### Where are the full trust reports for horovod and Megatron-LM? GraphCanon publishes per-repo trust reports with dated maintenance, provenance, and scan summaries: [horovod trust report](/tools/horovod-horovod/trust); [Megatron-LM trust report](/tools/nvidia-megatron-lm/trust). --- **Machine-readable endpoints** - JSON: [`/api/graphcanon/graph?tool=horovod-horovod`](/api/graphcanon/graph?tool=horovod-horovod) - LLM index: [/llms.txt](/llms.txt) - Full corpus: [/llms-full.txt](/llms-full.txt) _GraphCanon - The knowledge graph for AI development. https://www.graphcanon.com/_