Vercel AI Updates: October 18, 2026

1. AI Gateway Logs Add a Sessions View Showing How Session Affinity Routed Each Turn

Vercel. AI Gateway logs now group requests into sessions, with each row showing the client that ran it (such as Claude Code, Codex or Cursor), its cost, models, providers and the share of input served from cache across all turns. Opening a session shows its cache hit rate, duration and a per-turn path listing the model, timing and serving provider. Session affinity, which is on by default for text and chat requests, sends each turn to the provider that handled the previous one so more of the prompt is read from cache. Developers can group requests explicitly by sending the same sessionId in providerOptions.gateway or an x-session-id header; otherwise requests are grouped by a fingerprint of the conversation’s opening. Source

2. DeepSeek V4.1 Flash 1017 Snapshot Comes to AI Gateway

Vercel. AI Gateway added DeepSeek V4.1 Flash 1017 (deepseek/deepseek-v4.1-flash-1017), a minor update to DeepSeek V4.1 Flash. Vercel lists improvements in reasoning, terminal and repository-level coding, multi-step agentic tasks, and tool-assisted chart reading, along with shorter outputs that cut output tokens per agent run. The model is reachable through the AI SDK or the gateway’s OpenAI-compatible Chat Completions endpoint, with provider pricing passed through without markup. Source