Ce que Koyeb a publié en février 2026

Changelog #112: 50% Faster Time to Healthy for Deployments, Improved Deployments Using Scale-to-Zero, and more

Koyeb has improved deployment times by 30-50% through network propagation enhancements for Sandboxes and Web Services. This affects users who deploy applications on the platform, making their deployme…

Changelog

Koyeb has introduced several significant updates and new features to its platform. Notably, they have achieved a 50% faster time to healthy for deployments, improved deployments using Scale-to-Zero, a…

PaddlePaddle FastDeploy

Koyeb has introduced FastDeploy, an inference and deployment toolkit for large language models and visual language models based on PaddlePaddle. This toolkit offers production-ready deployment solutio…

Search code, repositories, users, issues, pull requests...

Koyeb has introduced a beta pre-release of their Model Context Protocol (MCP) server, which enables programmatic management of Koyeb resources such as apps, services, deployments, instances, and logs …

Koyeb Partner Hub

Koyeb has launched a Partner Hub to expand its ecosystem and support partners in scaling their businesses. The hub offers exclusive technical resources and support, allowing partners to stand out in t…

Changelog #111: Koyeb Enters Into A Definitive Agreement With Mistral, Rolling Out Manual Scaling of Services, and more

Koyeb has entered into a definitive agreement to join Mistral AI, bringing its platform, technology, and team to accelerate Mistral's Compute offering. This strategic move is designed to provide leadi…

Mistral Magistral Small 2506

Mistral AI has released the Magistral Small 2506 model, a 24-billion-parameter model built on Mistral Small 3, offering enhanced reasoning capabilities and multilingual support for dozens of languages…

DeepSeek-R1 0528 Qwen3 8B

Koyeb has launched one-click deployment for the DeepSeek-R1 0528 Qwen3 8B large language model, making it available on its high-performance cloud infrastructure. This change affects developers and bus…

Llama 4 Scout Instruct

Koyeb has launched one-click deployment for Llama 4 Scout Instruct, a 17 billion parameter large language model developed by Meta, on its GPU-powered cloud infrastructure. This change allows developer…

Mistral Devstral Small

Mistral AI has partnered with Koyeb to offer one-click deployment of its latest model, Mistral Devstral Small, on Koyeb's high-performance cloud infrastructure. This collaboration enables users to dep…

Qwen 3 Next 80B A3B Instruct

The most important change is that Koyeb now supports the deployment of Qwen 3 Next 80B A3B Instruct, a large language model from the Qwen3-Next series, on its cloud infrastructure. This affects develo…

Koyeb is Joining Mistral AI to Build the Future of AI Infrastructure

Koyeb, a serverless cloud infrastructure provider, has been acquired by Mistral AI to advance its AI infrastructure offerings, marking the most important development in this announcement. Koyeb's plat…

Run the GitHub Copilot CLI in Koyeb Sandboxes for Secure Code Generation of a Gradio App

GitHub has expanded the capabilities of GitHub Copilot by integrating it with Koyeb Sandboxes, allowing users to interact with Copilot directly from the command line in isolated, ephemeral environment…

Use Ollama to Test Multiple Code Generation Models With Koyeb Sandboxes

Koyeb has introduced Sandboxes, a new feature that allows developers to generate and execute AI-generated code in isolated, ephemeral cloud environments. This service addresses a critical challenge in…

Blog

Koyeb announced a definitive agreement to join Mistral AI, aiming to advance cutting-edge AI infrastructure. The deal signals a strategic shift for Koyeb, integrating its serverless platform with Mist…

GPU LLM Benchmarks

Koyeb released detailed GPU benchmarks for LLM inference across NVIDIA A100, H100, L40S, H200, B200, and RTX Pro 6000 GPUs, measuring latency, throughput, and token usage for models like Llama 3.1 8B,…

What is Koyeb?

Pre-Built Docker Images

How We Build from Git

Koyeb updated its Git-driven build and deployment documentation to clarify support for both Dockerfile and buildpack-based workflows. Users can now override Dockerfile settings, customize build comman…

Autoscaling

Koyeb introduced advanced autoscaling capabilities that dynamically adjust service instances based on CPU, memory, and request rates, with support for multiple scaling factors. The system scales up qu…

Managing Sandboxes with the Koyeb Python and JavaScript SDKs

Koyeb has introduced a new feature that allows users to create and manage isolated, ephemeral sandbox environments using their Python and JavaScript SDKs. This feature enables developers to safely run…

Scale-to-Zero

Koyeb added Scale-to-Zero to automatically scale Standard CPU and GPU instances to zero during idle periods, optimizing costs by charging only for active compute. Free-tier instances scale down after …

FLUX.1 Kontext [dev]

Koyeb has launched a one-click deployment for the FLUX.1 Kontext [dev] model, a 12 billion parameter AI model designed for text-based image editing, on its high-performance cloud infrastructure. This …

Changelog #110: Google Authentication Support for Koyeb Control Panel, Service Lifecycle Deletion Policy, and more

Koyeb has added Google Authentication support to its control panel, expanding login options beyond email/password and GitHub. This change affects all users, making it easier for teams to access their …

Qwen 3 Coder 30B A3B Instruct

Qwen and Koyeb have collaborated to deploy the Qwen 3 Coder 30B A3B Instruct model on Koyeb's GPU infrastructure, enabling high-performance, low-latency inference with autoscaling capabilities. This m…

Using the Claude Agent SDK with Koyeb Sandboxes for Secure Code Execution

The Claude Agent SDK now integrates with Koyeb Sandboxes to provide secure, isolated environments for executing AI-generated code, a critical advancement for safely deploying autonomous agents. This c…

Generate Code Using OpenAI Codex in a Koyeb Sandbox

Koyeb has introduced a tutorial for securely running OpenAI's Codex AI agent within its isolated sandbox environments. This change directly affects developers and engineering teams looking to integrat…

Koyeb Sandboxes: Fast, Scalable, Fully Isolated Environments for AI Agents and More

Koyeb has launched Koyeb Sandboxes in public preview, a new feature enabling developers to instantly spin up fully isolated, secure environments for executing AI-generated and untrusted code at scale.…

Koyeb Serverless GPUs: Slashing Prices on A100, H100, and L40S by up To 24%

Koyeb has cut prices by up to 24% on its high-performance serverless NVIDIA GPU instances, a significant move to make AI infrastructure more accessible and cost-competitive. This price reduction speci…

Using Dia 1.6B to Build a Text-to-Speech Application on Serverless GPUs

Koyeb has published a comprehensive tutorial for deploying a full-stack, production-ready text-to-speech (TTS) application using the Dia-1.6B model from Nari Labs, directly targeting developers and co…

Using PydanticAI to Build AI Agents for Data Analysis

The provided content is a detailed tutorial for building a financial reporting AI agent using PydanticAI, Ollama, and FastAPI, deployable on the Koyeb platform. This represents a significant competiti…

Serverless Postgres GA: Production-Ready Databases for Large Scale and AI Apps

The company has launched Serverless Postgres into general availability, a fully managed and auto-scaling database service now ready for production use, directly targeting developers building large-sca…

Partnering with Vultr for Serverless & Global AI Deployments

Koyeb has partnered with Vultr Cloud Alliance to enable global, serverless deployments of AI workloads across Vultr's 32 cloud regions. This integration directly affects developers and businesses usin…

DeepSeek-R1’s Multi-Lingual and Agentic RAG Capabilities in Practice

DeepSeek-R1 Qwen 32B, a major open-source AI model, has been successfully deployed and tested in production-grade agentic workflows, demonstrating competitive strengths in multilingual content generat…

Use OpenAI Apps SDK in ChatGPT with MCP deployed on Koyeb

OpenAI has released a tutorial demonstrating how to extend ChatGPT with a custom to-do list app using its Apps SDK and the Model Context Protocol (MCP), deployed on the Koyeb platform. This change pri…

Serverless GPUs: RTX Pro 6000, H200, and B200 Now Available on Koyeb

Koyeb has added three new high-performance NVIDIA GPUs—the RTX Pro 6000, H200, and B200—to its serverless platform, significantly expanding its competitive edge in on-demand AI infrastructure. This up…

The Infrastructure Behind Strapi Cloud: Serverless, High-Performance, Built on Koyeb

Strapi Cloud migrated its fully-managed hosting platform to Koyeb's serverless infrastructure to achieve superior performance, cost efficiency, and zero operational overhead. This technical partnershi…

Top Sandbox Platforms for AI Code Execution in 2026

Koyeb has positioned itself as the leading platform for securely executing AI-generated code in 2026, a critical competitive move as sandboxes become essential infrastructure for developers. The platf…

Serverless AI Infrastructure Going into 2026: Sandboxes, GPUs, and More

Koyeb's 2025 platform evolution centered on making production-grade AI deployment faster, more secure, and more cost-effective, directly addressing the industry's shift from prototypes to live applica…

Best Serverless GPU Platforms for AI Apps and Inference in 2026

Koyeb has launched a serverless GPU platform, positioning itself as a direct competitor to established providers like Modal and RunPod by offering global deployment, autoscaling, and support for high-…

Top PostgreSQL Database Free Tiers in 2026

Koyeb has published a competitive analysis of managed PostgreSQL databases with free tiers, positioning its own Database Service as a key contender. The article, aimed at developers and startups, eval…

Why You Should Run AI-Generated Code in a Sandbox

Koyeb has introduced sandboxed execution for AI-generated code, a critical security feature for developers and companies using AI coding agents. This change directly addresses the significant risks po…

Koyeb Sandboxes: JavaScript SDK Now Available

Koyeb has expanded its secure sandbox platform to JavaScript and TypeScript developers with the release of a new JavaScript SDK, directly competing in the fast-growing market for safe AI code executio…

Authentication: Announcing 2FA, MFA and Passkeys on Koyeb

Koyeb has launched a comprehensive overhaul of its authentication system, introducing critical security features like multi-factor authentication (MFA) and passkeys for all users. This update directly…

AI Dev Meetup on Coding Agents with OpenAI and LangChain

Koyeb hosted a major AI developer meetup in Paris focused on the competitive challenge of safely implementing AI coding agents. The event, which drew over 350 signups, featured technical talks from Ko…