Kimi K3.1 Pro Gains More ZDR Providers

This is the practical payoff of open weights: more competition at the serving layer, plus stronger privacy controls for teams deploying capable models.

• –Kimi K3.1 Pro combines a 2.8T-parameter MoE architecture, 104B active parameters, multimodal input, and a 1M-token context window.
• –ZDR routing makes the model more viable for sensitive coding, financial research, and agent workflows.
• –AI Gateway adds unified APIs, retries, failover, usage tracking, and provider routing around the model.
• –Highspeed targets interactive agents with faster responses, while the standard model remains cheaper.
• –Moonshot’s benchmark claims are ambitious, but independent testing across new providers will matter more than headline scores.