--- title: "cuvs vs cherche" type: "comparison" canonical_url: "https://www.graphcanon.com/compare/nvidia-cuvs-vs-raphaelsty-cherche" tools: ["nvidia-cuvs", "raphaelsty-cherche"] --- # cuvs vs cherche *GraphCanon updated Jul 12, 2026* ## Verdict Pick cuvs if cuVS is a CUDA-based library for efficient GPU-accelerated vector search and clustering; pick cherche if cherche is a Python library for implementing neural search capabilities. [cuvs](https://docs.rapids.ai/api/cuvs/stable/) reports 810 GitHub stars, 210 forks, and 645 open issues, last pushed Jul 11, 2026. [cherche](https://github.com/raphaelsty/cherche) has 331 stars, 14 forks, and 4 open issues, last pushed Jun 1, 2024. Figures are from public GitHub metadata via [cuvs's repository](https://github.com/NVIDIA/cuvs) and [cherche's repository](https://github.com/raphaelsty/cherche). | | [cuvs](/tools/nvidia-cuvs.md) | [cherche](/tools/raphaelsty-cherche.md) | | --- | --- | --- | | Tagline | A library for vector search and clustering on the GPU | Neural Search | | Stars | 810 | 331 | | Forks | 210 | 14 | | Open issues | 645 | 4 | | Language | Cuda | Python | | Adopt for | cuVS is a CUDA-based library for efficient GPU-accelerated vector search and clustering. | Cherche is a Python library for implementing neural search capabilities. | | Persona | - | - | | Runtime | - | - | | License | Apache-2.0 | MIT | | Categories | Vector Databases | Vector Databases, Data & Retrieval, Evaluation & Observability | ## Trust and health _Sourced signals - not a safety guarantee. No winner column._ | | [cuvs](/tools/nvidia-cuvs.md) | [cherche](/tools/raphaelsty-cherche.md) | | --- | --- | --- | | Maintenance | Very active (96%) | Dormant (18%) | | Days since push | 0d | 769d | | Open issues (now) | 645 | 4 | | Owner type | Organization | User | | Full report | [trust report](/tools/nvidia-cuvs/trust.md) | [trust report](/tools/raphaelsty-cherche/trust.md) | ## Decision facts: cuvs - **Adopt for:** cuVS is a CUDA-based library for efficient GPU-accelerated vector search and clustering. ## Decision facts: cherche - **Adopt for:** Cherche is a Python library for implementing neural search capabilities. ## Choose when ### Choose cuvs if… - cuvs is primarily Cuda; cherche is Python. - License: cuvs is Apache-2.0, cherche is MIT. - Tags unique to cuvs: clustering, anns, sparse, gpu. - cuvs ships Docker support for self-hosted deployment. - - When you need high-performance vector operations leveraging the parallel processing power of GPUs, specifically with CUDA. ### Choose cherche if… - cherche is primarily Python; cuvs is Cuda. - License: cherche is MIT, cuvs is Apache-2.0. - Tags unique to cherche: neural-networks, neural-search, nlp, machine-learning. - Also covers Data & Retrieval, Evaluation & Observability. - Cherche is a Python library for implementing neural search capabilities. ## When NOT to use cuvs - - For environments where GPU resources are limited or unavailable because cuVS heavily relies on CUDA's capabilities for performance gains. - - When you prioritize portability across different hardware, as cuVS being tied to CUDA means it may not be optimal on non-NVIDIA GPUs or CPU-only systems. ## When NOT to use cherche - Last GitHub push was 770 days ago (dormant maintenance, Jun 1, 2024). Validate activity before betting a new project on cherche. - Vector Databases: Don't reach for a dedicated vector DB under ~100k vectors; pgvector on your existing Postgres is simpler to operate. - Data & Retrieval: Skip a heavy ingestion framework when your corpus is small and static; a script plus the embedding API is enough. - Evaluation & Observability: Defer heavyweight eval infra only until you have real traffic - never skip it once users depend on answers. ## Common questions ### What is the difference between cuvs and cherche? cuvs: A library for vector search and clustering on the GPU. cherche: Neural Search. See the comparison table for live GitHub stats and shared categories. ### When should I choose cuvs over cherche? Choose cuvs over cherche when cuvs is primarily Cuda; cherche is Python; License: cuvs is Apache-2.0, cherche is MIT; Tags unique to cuvs: clustering, anns, sparse, gpu; cuvs ships Docker support for self-hosted deployment; - When you need high-performance vector operations leveraging the parallel processing power of GPUs, specifically with CUDA. ### When should I choose cherche over cuvs? Choose cherche over cuvs when cherche is primarily Python; cuvs is Cuda; License: cherche is MIT, cuvs is Apache-2.0; Tags unique to cherche: neural-networks, neural-search, nlp, machine-learning; Also covers Data & Retrieval, Evaluation & Observability; Cherche is a Python library for implementing neural search capabilities. ### When should I avoid cuvs? - For environments where GPU resources are limited or unavailable because cuVS heavily relies on CUDA's capabilities for performance gains. - When you prioritize portability across different hardware, as cuVS being tied to CUDA means it may not be optimal on non-NVIDIA GPUs or CPU-only systems. ### When should I avoid cherche? Last GitHub push was 770 days ago (dormant maintenance, Jun 1, 2024). Validate activity before betting a new project on cherche. Vector Databases: Don't reach for a dedicated vector DB under ~100k vectors; pgvector on your existing Postgres is simpler to operate. Data & Retrieval: Skip a heavy ingestion framework when your corpus is small and static; a script plus the embedding API is enough. Evaluation & Observability: Defer heavyweight eval infra only until you have real traffic - never skip it once users depend on answers. ### Is cuvs or cherche more popular on GitHub? cuvs has more GitHub stars (810 vs 331). Stars measure visibility, not whether either tool fits your constraints. ### Are cuvs and cherche open source? Yes - both are open-source projects on GitHub (cuvs: Apache-2.0, cherche: MIT). ### Where can I find alternatives to cuvs or cherche? GraphCanon lists graph-backed alternatives at [cuvs alternatives](/tools/nvidia-cuvs/alternatives) and [cherche alternatives](/tools/raphaelsty-cherche/alternatives) ([cuvs markdown twin](/tools/nvidia-cuvs/alternatives.md), [cherche markdown twin](/tools/raphaelsty-cherche/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/nvidia-cuvs-vs-raphaelsty-cherche.md) mirrors this page for agents and LLM crawlers, with the same stats table and FAQ answers. ### Which is better maintained, cuvs or cherche? cuvs: Very active. cherche: 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 cuvs and cherche? GraphCanon publishes per-repo trust reports with dated maintenance, provenance, and scan summaries: [cuvs trust report](/tools/nvidia-cuvs/trust); [cherche trust report](/tools/raphaelsty-cherche/trust). --- **Machine-readable endpoints** - JSON: [`/api/graphcanon/graph?tool=nvidia-cuvs`](/api/graphcanon/graph?tool=nvidia-cuvs) - 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/_