← Back to live feed1 story
A new integration brings multimodal Mixture-of-Experts capabilities to developers, providing a 1M token context window for large-scale projects. OpenCode Go now supports Alibaba's Qwen3.8-Flash, which features 125B total parameters with 6B activated per token to allow high-capacity reasoning with a reduced compute footprint.
The release serves as a technical preview of the Qwen4 architecture, emphasizing efficiency in coding and office tasks. The model is currently available through OpenCode Go, QwenCloud and OpenRouter.
Sign in to suggest edits