Build AI models with Tenstorrent

Access open, scalable, and powerful next-gen AI hardware in seconds — Get started free

Tenstorrent

Koyeb is trusted by the most ambitious teams

Develop, Test and Run AI Models on Tenstorrent
Tenstorrent’s open-source ecosystem and RISC-V architecture offer a powerful, customizable, and scalable environment for building, optimizing, and deploying AI models efficiently. From single-user/single-model to multi-user/multi-model environments, optimize inference performance to meet your needs.
Offering

On-demand Tenstorrent access

Koyeb provides the serverless platform to instantly access Tenstorrent hardware.

Autoscaling
full-fledged containers

Autoscaling full-fledged containers

Spin up both development and production environments with zero configuration. We automatically scale your resources up with demand and down to zero when traffic drops.
Instant development environment

Instant development environment

Deploy ready-to-use development environment with just one click. Connect Tenstorrent instances via VSCode, run Docker containers, and more—all with zero configuration required.
Everything you need to build and run on Tenstorrent

Everything you need to build and run on Tenstorrent

Access TT-Metalium to write host and kernel programs for tasks like matrix multiplication or image resizing. Build and use libraries as building blocks like TT-NN to get flexible high-level environment and run a variety of ML applications.
Scale on high-performance infrastructure

Scale on high-performance infrastructure

Combine Koyeb's serverless, AI-optimized inference platform with Tenstorrent accelerators for a streamlined, efficient solution. Run AI models and ML applications without worrying about infrastructure management.
Wormhole™ PCIe Boards: RISC-V-Powered AI Acceleration for Less Than GPUs
Wormhole™ PCIe boards, powered by Tensix Cores, combine a compute unit, network-on-chip, local cache, and “baby RISC-V” cores for efficient data movement and high-performance AI processing. With superior performance-to-cost ratios over traditional GPUs and support for broad data precision formats (FP8, FP16, BF16, FP32), Wormhole™ is optimized for scalable AI workloads, from training to inference at scale.
Wormhole board
/What people are saying

On demand access enables developers to experience Tensix processors in a minute - allowing them to develop faster. Cloud access is important because we want developers to be able to try our technology and see that we are serious about their experience and their feedback before committing to ownership. This is one of the ways we enable them to own their future.

Jim Keller
Jim Keller
CEO
View customer stories
AI Compute Fabric: Scalable, High-Performance Cloud AI

Wormhole™ PCIe boards create a scalable AI compute fabric, tailored for cloud-native workloads. Featuring up to 96GB of on-chip RAM and high-speed interconnects, they excel in large-scale training and inference.

Achieve real-time inference, ensure zero-downtime deployments, and dynamically scale resources for optimal performance and cost-efficiency.

Offering

Everything you need for production

Instant API Endpoint
Instant API Endpoint

Just hit deploy to provision an API endpoint ready to handle requests in seconds. No waiting, no config.

Get Started
Smart and Fast Autoscaling
Smart and Fast Autoscaling

Get efficient autoscaling to adapt infrastructure to demand, with imperceptible cold start.

Learn more
GPU, NPU, Accelerators, or just CPU
GPU, NPU, Accelerators, or just CPU

Access a wide range of optimized hardware for scale-out workloads, on demand, in seconds.

Deploy now
Zero-Downtime Deployments
Zero-Downtime Deployments

Enjoy built-in continuous deployment with automatic health checks to prevent bad deployment and ensure you’re always up and running.

View docs
Native HTTP/2, WebSocket, and gRPC Support
Native HTTP/2, WebSocket, and gRPC Support

Stream large or partial responses to end-users and accelerate your connections through a global edge network for instant feedback and responsive applications.

Get started
/and much more
Ultra-Fast NVME Storage

Store datasets, models, and fine-tune weights on blazing-fast NVME disks offering extremely high write and read throughput for exceptional performance.

Get started
Logs and Instance Access

Troubleshoot and investigate issues easily using real-time logs, or directly connect to your instances.

Get started
Private VPC for Micro-Services

Secure your service-to-service communications with a built-in, ops-free, service mesh. The private network is end-to-end encrypted and authenticated with mTLS.

Get started

Ready to build on Tenstorrent hardware?

Koyeb is a developer-friendly serverless platform to deploy apps globally. No-ops, servers, or infrastructure management.
All systems operational
© Koyeb