This week marks our 100th changelog, a milestone worth celebrating! While we’ve been building far longer than 100 weeks (shh 
), each installment represents the steady evolution of the Koyeb platform, guided by your feedback and powered by our incredible developer community.
From introducing new hardware and speeding up deploys, to optimizing autoscaling and launching features like Scale-to-Zero and Light Sleep (for those lightning-fast Instance wakeups under 200 ms 
), we’ve come a long way. With one-click apps and constant improvements, Koyeb is faster, more powerful, and more fun than ever to build on.
Here’s to the next hundred! 
1. 2× Faster startup time for GPU-Backed Services
GPU instance allocation is now 2× faster, cutting startup times in half. Complete your deploys, redeploys, and scale-ups in half the time for faster rollouts, quicker feedback loops, and a smoother experience on GPU-backed services. Deploy now on GPU.
2. Quickly access and manage your resources with the new command palette
Taking a page from your favorite IDEs, Koyeb’s command palette has received a full revamp. The palette now adapts to your current context, allowing you perform the most relevant actions right from your keyboard and comes with a bunch of new actions. Open the palette with Ctrl+K on Linux/Windows or Cmd+K on Mac. See something missing? Let us know what actions you’d like next.
3. Koyeb at All Things Open 2025
Next week, Koyeb Co-founder the CEO Yann Leger will be speaking alongside Greg Wallace, Director of Partnerships at NetActuate about AI serverless infrastructure on Tenstorrent hardware at All Things Open. Come join us to learn about the evolution and future of AI infrastructure. Get the full event details here and the talk details here.
4. New One-Click Qwen 3 VL 235B A22B, blog on building an AI-powered photo booth
Part of the Qwen 3 VL family, Qwen 3 VL 235B A22B Instruct is an advanced open-source vision-language model. You can deploy Qwen 3 VL on Koyeb with one click and take advantage of this model’s superior text understanding and generation, deeper visual perception and reasoning, and more. Read the blog to see how we used this model in our AI Engineer Paris photo booth.
