Nebul Adds Support for Alibaba’s Qwen3 Models

We’ve expanded our Private Inference API and PrivateGPT offerings with the addition of Alibaba’s latest large language models, the Qwen3 family. These models represent a significant step forward from their previous QwQ and Qwen2.5 iterations, offering notable advancements in several key areas.

Qwen3 features a flexible architecture that allows for dynamic switching between a deliberate “thinking mode” for complex reasoning and a fast “non-thinking mode” for efficient general chat. The models also provide extensive multilingual support, covering over 100 languages and dialects, and demonstrate improved capabilities for integrating with external tools for agent-based applications.

Our team has been evaluating the performance of the Qwen3 models, particularly the 32B variant, across various GPU configurations over the past few days. Initial results indicate strong performance in both throughput and the quality of generated output.

Qwen3 is now available on our platform, providing users with access to advanced AI capabilities within a private and secure environment.

For more information or to schedule a demonstration, feel free to get in touch with us!