Skip to content
NO FEED

Understand the changes. Choose the tools. Put AI to work.

← News
News

DeepSeek V4 Pro leaves preview as 0813; Grok 4.6 ships the same day for long-running agents in Cursor and Grok Build.

DeepSeek's pricing page now lists deepseek-v4-pro as DeepSeek-V4-Pro-0813 — GA after a preview since April.1 Same model id, OpenAI and Anthropic base URLs, tool calls, Responses API, 1M context. Cache-miss input $0.435 / 1M, output $0.87 / 1M; a footnote says a significant API price rise is planned, plan still "subject to official notice."1 Vercel AI Gateway routes deepseek/deepseek-v4-pro to the new weights by default; pin deepseek/deepseek-v4-pro-0813 to be explicit.2

SpaceXAI released Grok 4.6 the same window: long-running agents, live in Cursor and Grok Build with 2× included usage the first week.34 API id grok-4.6, 500k context, $2 / $6 per 1M in/out, reasoning up to xhigh.5 Vendor chart ties it with GPT-5.6 Sol Max at 61 on the AA Intelligence Index (Fable 5 Max 62). Time your own harness before you move the default.

Use it

Pin Pro 0813 and run one trusted agent loop; budget for a price change with no date. Try Grok 4.6 on one long job this week under the 2× promo; keep Composer for cheap daily edits.

Deeper — pins and limits
  • Flash → Flash-0731, Pro → Pro-0813; concurrency 2500 / 500.1
  • No independent public bench of 0813 opened this morning.
  • Grok 4.6 also on OpenRouter, Vercel, Cloudflare; fast variant is 2× list price.35
  • Cursor: desktop, cloud agents, iOS, CLI, SDK; 4.5 stays.4

Sources

  1. DeepSeek docs — Models & Pricing (V4-Pro-0813) · read
  2. Vercel changelog — DeepSeek V4 Pro weights on AI Gateway · read
  3. x.ai — Introducing Grok 4.6 · read
  4. Cursor forum — Grok 4.6 is now Live · read
  5. xAI docs — grok-4.6 model card · read
Originally part of the 13 August 2026 edition.

The useful part, in your inbox.

One complete edition every weekday, with practical takeaways.