--- title: "ray-llm vs exllama" type: "comparison" canonical_url: "https://www.graphcanon.com/compare/ray-project-ray-llm-vs-turboderp-exllama" tools: ["ray-project-ray-llm", "turboderp-exllama"] --- # ray-llm vs exllama *GraphCanon updated Sep 20, 2026* ## Verdict Pick ray-llm if ray-LLM is an archived project for deploying large language models on Ray, with its functionalities now integrated into the Ray framework through the ray.serve.llm and ray.data.llm APIs; pick exllama if exLlama provides a memory-efficient implementation of the LLaMa model with support for quantized weights, primarily aimed at users with NVIDIA GPUs from the 30-series onwards. [ray-llm](https://docs.ray.io/en/latest/) reports 1.3k GitHub stars, 90 forks, and 0 open issues, last pushed Mar 13, 2025. [exllama](https://github.com/turboderp/exllama) has 2.9k stars, 220 forks, and 65 open issues, last pushed Sep 30, 2023. Figures are from public GitHub metadata via [ray-llm's repository](https://github.com/ray-project/ray-llm) and [exllama's repository](https://github.com/turboderp/exllama). | | [ray-llm](/tools/ray-project-ray-llm.md) | [exllama](/tools/turboderp-exllama.md) | | --- | --- | --- | | Tagline | Archived repository for LLMs on Ray, now integrated into Ray. | Memory-efficient rewrite of HF transformers for Llama with quantized weights | | Stars | 1,261 | 2,943 | | Forks | 90 | 220 | | Open issues | 0 | 65 | | Language | - | Python | | Adopt for | Ray-LLM is an archived project for deploying large language models on Ray, with its functionalities now integrated into the Ray framework through the ray.serve.llm and ray.data.llm APIs. | ExLlama provides a memory-efficient implementation of the LLaMa model with support for quantized weights, primarily aimed at users with NVIDIA GPUs from the 30-series onwards. | | Persona | - | - | | Runtime | - | - | | License | License information is not provided. | MIT | | Categories | Inference & Serving, Model Training | Inference & Serving, LLM Frameworks | ## Trust and health _Sourced signals - not a safety guarantee. No winner column._ | | [ray-llm](/tools/ray-project-ray-llm.md) | [exllama](/tools/turboderp-exllama.md) | | --- | --- | --- | | Maintenance | Archived (8%) | Dormant (18%) | | Days since push | 554d | 1072d | | Archived on GitHub | Yes | No | | Open issues (now) | 0 | 65 | | Stars delta | 0 (30d) | +6 (30d) | | Owner type | Organization | User | | Full report | [trust report](/tools/ray-project-ray-llm/trust.md) | [trust report](/tools/turboderp-exllama/trust.md) | ## Decision facts: ray-llm - **Pricing:** unknown - Pricing details are not available. - **Requirements:** Ensure you have the latest version of Ray installed to use the ray.serve.llm and ray.data.llm APIs. - **Adopt for:** Ray-LLM is an archived project for deploying large language models on Ray, with its functionalities now integrated into the Ray framework through the ray.serve.llm and ray.data.llm APIs. - **License detail:** License information is not provided. ## Decision facts: exllama - **Adopt for:** ExLlama provides a memory-efficient implementation of the LLaMa model with support for quantized weights, primarily aimed at users with NVIDIA GPUs from the 30-series onwards. ## Choose when ### Choose ray-llm if… - Pricing: Pricing details are not available.. - Requirements: Ensure you have the latest version of Ray installed to use the ray.serve.llm and ray.data.llm APIs.. - Tags unique to ray-llm: llm, llm-serving, ray. - Also covers Model Training. - If you are working with Ray and need to deploy large language models, use the ray.serve.llm and ray.data.llm APIs, which are now integrated into the Ray framework. ### Choose exllama if… - Tags unique to exllama: docker, llama model, memory-efficient, nvidia gpu. - Also covers LLM Frameworks. - exllama ships Docker support for self-hosted deployment. - - When deploying LLaMa models on NVIDIA GPUs from the 30-series or later that have strong FP16 support. ## When NOT to use ray-llm - Do not use the archived Ray-LLM repository for active development or deployment, as it is no longer maintained and its functionalities have been integrated into the Ray framework. - If you require the latest features and support for deploying large language models, avoid the archived repository and use the ray.serve.llm and ray.data.llm APIs directly. ## When NOT to use exllama - - If you are operating older GPUs such as Pascal series, which lack robust FP16 support; alternatives like AutoGPTQ might perform better. - - In scenarios that involve AMD GPU hardware (due to limited testing and optimization efforts). ## Common questions ### What is the difference between ray-llm and exllama? ray-llm: Archived repository for LLMs on Ray, now integrated into Ray.. exllama: Memory-efficient rewrite of HF transformers for Llama with quantized weights. See the comparison table for live GitHub stats and shared categories. ### When should I choose ray-llm over exllama? Choose ray-llm over exllama when Pricing: Pricing details are not available.; Requirements: Ensure you have the latest version of Ray installed to use the ray.serve.llm and ray.data.llm APIs.; Tags unique to ray-llm: llm, llm-serving, ray; Also covers Model Training; If you are working with Ray and need to deploy large language models, use the ray.serve.llm and ray.data.llm APIs, which are now integrated into the Ray framework. ### When should I choose exllama over ray-llm? Choose exllama over ray-llm when Tags unique to exllama: docker, llama model, memory-efficient, nvidia gpu; Also covers LLM Frameworks; exllama ships Docker support for self-hosted deployment; - When deploying LLaMa models on NVIDIA GPUs from the 30-series or later that have strong FP16 support. ### When should I avoid ray-llm? Do not use the archived Ray-LLM repository for active development or deployment, as it is no longer maintained and its functionalities have been integrated into the Ray framework. If you require the latest features and support for deploying large language models, avoid the archived repository and use the ray.serve.llm and ray.data.llm APIs directly. ### When should I avoid exllama? - If you are operating older GPUs such as Pascal series, which lack robust FP16 support; alternatives like AutoGPTQ might perform better. - In scenarios that involve AMD GPU hardware (due to limited testing and optimization efforts). ### Is ray-llm or exllama more popular on GitHub? exllama has more GitHub stars (2,943 vs 1,261). Stars measure visibility, not whether either tool fits your constraints. ### Are ray-llm and exllama open source? Yes - both are open-source projects on GitHub. ### Where can I find alternatives to ray-llm or exllama? GraphCanon lists graph-backed alternatives at [ray-llm alternatives](/tools/ray-project-ray-llm/alternatives) and [exllama alternatives](/tools/turboderp-exllama/alternatives) ([ray-llm markdown twin](/tools/ray-project-ray-llm/alternatives.md), [exllama markdown twin](/tools/turboderp-exllama/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/ray-project-ray-llm-vs-turboderp-exllama.md) mirrors this page for agents and LLM crawlers, with the same stats table and FAQ answers. ### Which is better maintained, ray-llm or exllama? ray-llm: Archived. exllama: Dormant. 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 ray-llm and exllama? GraphCanon publishes per-repo trust reports with dated maintenance, provenance, and scan summaries: [ray-llm trust report](/tools/ray-project-ray-llm/trust); [exllama trust report](/tools/turboderp-exllama/trust). --- **Machine-readable endpoints** - JSON: [`/api/graphcanon/graph?tool=ray-project-ray-llm`](/api/graphcanon/graph?tool=ray-project-ray-llm) - 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/_