Open Brain
VercelEngineering PostOfficial Source

Kimi K3.1 Pro now has more ZDR providers on AI Gateway

Kimi K3.1 Pro is now available through more ZDR providers on AI Gateway. Builders can route coding agents to Pro or Highspeed while requiring requests to use ZDR-backed providers.

Vercel · Nov 4, 2026
Open Source Open MarkdownOpen JSON
Source Summary

Kimi K3.1 Pro and **K3.1 Pro Highspeed** are available through more Zero Data Retention providers on AI Gateway; previously, only Moonshot AI served them. Both accept images and have a **1M-token context window**. Highspeed is aimed at interactive use such as coding agents.

Practical Implication

You can connect Claude Code, Codex, Cursor, or another supported client through AI Gateway. Set `zeroDataRetention` per request or enable ZDR in dashboard settings to require ZDR routing across requests, including agent traffic.

Agent-Ready Context
Kimi K3.1 Pro and **K3.1 Pro Highspeed** are available through more Zero Data Retention providers on AI Gateway; previously, only Moonshot AI served them. Both accept images and have a **1M-token context window**. Highspeed is aimed at interactive use such as coding agents.

You can connect Claude Code, Codex, Cursor, or another supported client through AI Gateway. Set `zeroDataRetention` per request or enable ZDR in dashboard settings to require ZDR routing across requests, including agent traffic.

If no ZDR provider is available for a model, the request fails instead of routing to a non-ZDR provider. The material gives no details on provider names, latency, or comparative model performance.
Connected Context · Feed7 Judgment

This adds a data-retention and routing dimension to the earlier Kimi K3.1 Pro listing: more ZDR providers can serve both variants, and requests fail if none is available rather than falling back to non-ZDR routing. It does not establish provider names, latency, or comparative model quality.

Context Map
modelcoding#model-selection#coding-agents
Uncertainty
If no ZDR provider is available for a model, the request fails instead of routing to a non-ZDR provider. The material gives no details on provider names, latency, or comparative model performance.