Blog
Read about our latest announcements.Claude Agent SDK with ZenMux: Cut Agent Loop Costs, Keep Loops Running
Point the Claude Agent SDK at ZenMux to access 200+ models, route phases of your agent loop to cheaper models, and keep loops running through provider outages — all behind one API key.
An Experimental Validation Plan for the Feedback That "ZenMux Has Worse Cache Hit Rates Than OpenRouter"
This article describes an experimental framework for comparing cache hit rates between ZenMux and OpenRouter. By fixing the model, provider, and prompt sequence, and by measuring cached_tokens and token_hit_rate across multi-turn conversations, the experiment shows that there is no significant difference in cache hit rate between the two platforms under the tested conditions.
Debugging a Claude Code Freeze
Recently, several users reported that Claude Code would freeze when calling models through ZenMux — after sending a prompt, nothing appeared on screen, and the only option was to force quit. As a model aggregation platform, we needed to figure out whether this was a ZenMux issue or something else entirely.
ZenMux User Agreement Update Notice
ZenMux User Agreement Update Notice
Billing Anomaly Notice and Compensation
Billing Anomaly Notice & Compensation
Nano Banana Pro Now Available on ZenMux: Next-Generation Image Intelligence Powered by Gemini 3 Pro
Experience Nano Banana Pro on ZenMux—Google’s next-gen Gemini 3 Pro image model with 4K resolution, faster processing, and enhanced reasoning. Free version is also available to use now.
Gemini 3 Pro Preview Now Available on ZenMux: Multimodal AI via a Unified API
Gemini 3 Pro Preview, Google’s most advanced multimodal reasoning model, is now available on ZenMux. Through ZenMux’s unified API, developers can access frontier‑level text, image, audio, video, and code understanding with enterprise‑grade reliability, and seamless integration alongside other leading models.
Ring-1T Now Available on ZenMux: Trillion-Parameter Thinking AI for Developers and Enterprises
Discover Ring-1T on ZenMux, a trillion-parameter thinking AI by Ant Group inclusionAI. Experience advanced reasoning, code generation, and math problem solving with 128K context, stable RL training, and enterprise-ready API integration.
Ling-1T is Now Available on ZenMux: Trillion-Scale Reasoning Power for Developers
Discover Ling-1T on ZenMux — a trillion-parameter reasoning model with 128K context, SOTA efficiency, and affordable API access for developers worldwide.