# Cohere Eliminates Storage and GPU Bottlenecks at Scale with NeuralMesh

**Customer:** Cohere

Cohere works with WEKA® NeuralMesh™ to meet the demands of massive training and inference workloads across multicloud environments.

## Results

- **10x** faster checkpointing

[Cohere](https://cohere.com/) is the driving force behind AI innovation for some of the world’s largest organizations. They are pioneering the creation of enterprise-grade AI and large language models built for privacy, security, and scale.

NeuralMesh delivers everything they need: unmatched performance, efficiency, and the ability to eliminate storage and GPU bottlenecks at scale.

> Cohere runs on NeuralMesh to achieve 10x faster checkpointing, accelerated read/write throughput, and best-in-class performance at lower cost for our AI model training and inference clusters. It delivers without breaking a sweat.
>
> — Autumn Moulder, Director of Infrastructure & Security • Cohere