SHELFCRITTER AI Digest

MODEL

Grok 4.6

modeltopic-notexaispacex

Overview

Grok 4.6 is xAI‘s August 2026 frontier release — the successor to Grok 4.5 and the first xAI model to explicitly move the competitive frame from “which model is best” to “which model is cheap enough to route the median agent call to.” Shipped 2026-08-12 with an Artificial Analysis Intelligence Index of 61 (tying GPT-5.6 Sol, behind Claude Opus 5) and a GDPval-AA v2 Elo of 1,753 (second overall). Headline pricing is $2 / $6 per M input/output tokens for short context, doubling to $4 / $12 above the 200K-token band — a load-bearing distinction that reshapes the “60%+ cheaper than Claude Opus 5 ($5/$25) and GPT-5.6 Sol ($5/$30)” story into cheaper on the workload most agent traffic sits in, not a flat undercut. Distribution shipped simultaneously on xAI API, Cursor, Grok Build, OpenRouter, Vercel, and Cloudflare.

Timeline

  • 2026-08-13-AI-Digest — xAI released Grok 4.6 on 2026-08-12 with an Artificial Analysis Intelligence Index of 61 (tying GPT-5.6 Sol, behind Claude Opus 5) and a GDPval-AA v2 Elo of 1,753 (second overall). Headline pricing $2 / $6 per M input/output at short context, doubling to $4 / $12 above 200K tokens — the long-context step is the load-bearing detail. Distribution shipped simultaneously on xAI API, Cursor, Grok Build, OpenRouter, Vercel, and Cloudflare. Narrow read the corpus carries: “60%+ cheaper than Claude Opus 5 / GPT-5.6 Sol” holds ONLY at short context — above 200K tokens the delta compresses sharply; frame as cheaper on the workload most agent traffic sits in, not a flat undercut. Structural read: three frontier-or-adjacent drops this week — Grok 4.6, DeepSeek V4 Pro 0813, Muse Glimmer — are all pricing or distributing to undercut the Anthropic / OpenAI price bracket rather than beat them on a headline benchmark. The competitive front is moving from which model is best to which model is cheap enough to route the median agent call to. 30 / 60 / 90-day watch: whether Anthropic and OpenAI respond with cache-write / batch discount refreshes rather than headline rate cuts; whether Grok 4.6 lands on the Aider polyglot leaderboard; whether the short-vs-long-context pricing split becomes the default frontier-tier pricing shape.

Key Developments

  1. Frontier-Tier Price / Perf Move at $2 / $6 Short Context (August 12, 2026): Grok 4.6 ties GPT-5.6 Sol on the Artificial Analysis Intelligence Index (both at 61) at 60%+ lower short-context pricing than Claude Opus 5 ($5/$25) and GPT-5.6 Sol ($5/$30). The load-bearing framing to carry: the undercut is short-context-specific — above 200K tokens the rate doubles to $4 / $12, and the delta compresses sharply. This is cheaper where median agent traffic lives, not a flat undercut. GDPval-AA v2 Elo of 1,753 (second overall) is the corroborating capability datapoint alongside the AA Intelligence Index tie.

  2. Same-Day Distribution Across Six Surfaces (August 12, 2026): xAI API, Cursor, Grok Build, OpenRouter, Vercel, and Cloudflare all shipped Grok 4.6 access on release day — a distribution simultaneity that reads as continuation of the Grok 4.5 / Cursor integration story (xAI-inside-SpaceX-inside-Cursor from the June 2026 acquisition), and positions Grok 4.6 as immediately routable rather than gated behind partner-preview windows.

  3. Third Open-or-Adjacent Frontier Drop in Three Days (August 12, 2026): Grok 4.6 lands the same day as DeepSeek V4 Pro‘s stealth 0813 checkpoint on OpenRouter and two days after Muse Glimmer (Aug 10). The corpus framing: three frontier-or-adjacent drops in three days, all pricing or distributing to undercut the Anthropic / OpenAI bracket rather than beat them on a headline benchmark — the competitive front is moving from capability-first to price-per-workflow-token-first.

See also: xAI, Grok 4.5, Cursor, SpaceX, GPT-5.6 Sol, Claude Opus 5, DeepSeek V4 Pro, Muse Glimmer, MOC - Major Companies, MOC - Open Source Models.