xAI’s Grok 4.6 is now rolling out in GitHub Copilot, two days after the mannequin’s August 12, 2026 launch, the firm introduced on August 14, 2026. The mannequin — xAI’s present coding flagship, pitched at long-running brokers and multi-step work — is selectable from Copilot’s mannequin picker throughout eight surfaces: VS Code, Visible Studio, the Copilot CLI, the Copilot cloud agent, the Copilot app, JetBrains IDEs, Xcode, and Eclipse.
GitHub’s changelog entry, revealed the identical day, confirms the rollout covers Copilot Professional, Professional+, Max, Enterprise, and Enterprise plans. That unfold is unusually broad for a brand new mannequin addition: JetBrains, Xcode, and Eclipse assist places Grok 4.6 in entrance of builders nicely exterior the VS Code core the place most Copilot mannequin launches focus.
GitHub mentioned that in its inside testing, Grok 4.6 carried out particularly nicely on longer-horizon duties requiring sustained reasoning and power use throughout terminal-based coding in VS Code and the Copilot CLI. That aligns with how xAI framed the mannequin at launch. “Grok 4.6 builds on Grok 4.5 with a specific concentrate on long-running brokers and extra formidable interactive and visible work,” the corporate wrote in its launch put up.
What Grok 4.6 Brings to the Mannequin Picker
Grok 4.6 arrives in Copilot with a set of vendor-reported analysis outcomes that place it close to the highest of the present coding class, with the caveat that applies to all self-reported numbers: these are xAI’s runs, on xAI’s chosen harnesses. Within the firm’s revealed eval desk, Grok 4.6 Excessive scores 61 on the Synthetic Evaluation Intelligence Index, a composite of 9 benchmarks, matching OpenAI’s GPT-5.6 Sol Max and sitting one level behind Anthropic’s Fable 5 Max at 62. On GDPVal-AA, a knowledge-work analysis, Grok 4.6 posts 1753 in opposition to 1728 for GPT-5.6 Sol Max and 1741 for Fable 5 Max.
The unfold is wider on agentic coding benchmarks. Grok 4.6 scores 69.9% on CursorBench v3.2, forward of GPT-5.6 Sol Max at 67.2% and behind Fable 5 Max at 70.5%. On DeepSWE v1.1 it reaches 65.9%, nicely forward of its predecessor Grok 4.5 Excessive at 54% however behind GPT-5.6 Sol Max’s 73%. On Terminal-Bench v3.0, which measures precisely the type of terminal work GitHub highlighted, Grok 4.6 scores 26% in opposition to 34.6% for GPT-5.6 Sol Max, whereas nonetheless practically doubling Grok 4.5 Excessive’s 15.7%.
The mannequin’s coaching recipe, per xAI’s account, paired an extended supplemental coaching run on curated reasoning and engineering information with supervised fine-tuning trajectories regenerated by Grok 4.5, adopted by reinforcement studying throughout agentic duties together with kernel optimization, net improvement, and computer-aided design. The generation-over-generation bounce on terminal and agentic benchmarks is the seen outcome, even when Grok 4.6 doesn’t lead the class on each measure.
The Superb Print on Entry and Billing
Two constraints form who really sees the mannequin and what it prices. First, the rollout is gradual: GitHub notes that builders who don’t see Grok 4.6 within the picker but ought to test again as availability expands. Second, on Copilot Enterprise and Enterprise plans the Grok 4.6 coverage is off by default, and an administrator should allow it in Copilot settings earlier than seats can choose the mannequin.
On billing, Grok 4.6 is charged at supplier record pricing underneath usage-based billing somewhat than drawing down a flat request allowance. Grok 4.6 is billed at supplier record pricing; xAI lists API pricing beginning at $2 per million enter tokens and $6 per million output tokens, the identical base price GitHub prices for Grok 4.5 in its pricing desk, and it places Grok 4.6 nicely beneath the opposite flagship choices within the picker on value: GPT-5.6 Sol lists at $5 enter and $30 output per million tokens, and Claude Opus-class fashions at $5 and $25.
Exterior Copilot, Grok 4.6 stays accessible by way of xAI’s API, Cursor, Grok Construct, and companions together with OpenRouter, Vercel, and Cloudflare. As coated in Unite.AI’s report on the launch, xAI supplied doubled included utilization in Grok Construct and Cursor for the mannequin’s first week, a window that closes August 19, 2026.
The Copilot integration extends a distribution sample that has outlined Grok 4.6’s first 48 hours: launch in xAI’s personal tooling and Cursor, then arrive inside the biggest third-party coding assistant by seat rely. With the Enterprise and Enterprise coverage gate now in directors’ palms, the tempo of the mannequin’s enterprise uptake will present up in Copilot settings earlier than it exhibits up in any leaderboard.
