Kimi K3.1 Pro now has more ZDR providers on AI Gateway
Vercel’s AI Gateway now supports Kimi K3.1 Pro and K3.1 Pro Highspeed from additional Zero Data Retention providers beyond Moonshot AI, expanding access for coding, 3D modeling, and financial analysis workflows.
Vercel’s AI Gateway has expanded support for Kimi K3.1 Pro and its Highspeed variant to include more Zero Data Retention providers, previously limited to Moonshot AI. Both models feature a 1M-token context window and image input capabilities, with Highspeed optimized for interactive tasks like coding agents where response time is critical. Users can call either model through AI Gateway’s SDK, Chat Completions, Responses, or Anthropic Messages APIs, or via coding agents connected to the platform.
To enforce Zero Data Retention, developers must set the zeroDataRetention option in the gateway provider configuration or enable it globally in the AI Gateway dashboard. When enabled, AI Gateway routes requests exclusively to providers with a ZDR agreement; if no compliant provider is available, the request fails rather than defaulting to a non-ZDR provider. This ensures compliance with data retention policies for all team requests, including those from coding agents.
Integration with popular coding agents such as Claude Code, Codex, and Cursor is now supported by installing the latest Vercel CLI and running the setup command. Agents can then select the appropriate model—moonshotai/kimi-k3.1-pro or moonshotai/kimi-k3.1-pro-highspeed—directly within their workflows. Vercel provides a guide for configuring coding agents to use these models via AI Gateway.
AI Gateway serves as a unified interface for model calls, offering features like usage tracking, cost management, retries, failover, and performance optimizations to improve uptime. The platform includes built-in reporting, Zero Data Retention support, budget controls for API keys, and routing rules. Pricing aligns with provider rates with no markup, and no platform fee is charged on inference, including Bring Your Own Key requests. A full list of available language models is accessible through the platform.