--- title: "GitHub Copilot Bolts On Grok 4.5 While Microsoft Promises to Someday Unify the Mess" date: 2026-07-30 tags: ["github-copilot","microsoft","grok","cursor","agentic-workflows"] categories: ["AI Tools"] summary: "GitHub added xAI's Grok 4.5 to Copilot on July 28 — a 500K-context model bolted onto an already multi-model picker. A day later, Satya Nadella told investors Microsoft is building one unified app to merge Copilot Chat, GitHub Copilot, Copilot Cowork, and Autopilot, sometime in 2026. That roadmap gap is the whole story." --- ![GitHub Copilot Bolts On Grok 4.5 While Microsoft Promises to Someday Unify the Mess](/images/github-copilot-grok-4-5-unified-super-app.png) Two Microsoft announcements landed a day apart this week, and read together they say more than either does alone. On July 28, GitHub's changelog quietly added a fifth-ish model — xAI's Grok 4.5 — to the already-crowded Copilot model picker. On July 29, on Microsoft's earnings call, Satya Nadella told investors the company is building a single app to finally merge Copilot Chat, GitHub Copilot, Copilot Cowork, and the Autopilot agent system. One announcement adds another piece to a fragmented product line; the other admits, on the record, that the fragmentation is a problem worth fixing. Claude Code shipped as one coherent terminal-native tool from day one and never needed either move. ## What actually shipped Grok 4.5 became selectable in GitHub Copilot starting July 28, rolling out across VS Code, JetBrains IDEs, Xcode, and Copilot's other surfaces. The model brings a 500,000-token context window and three selectable reasoning-effort tiers (low, medium, high), positioned for agentic coding tasks. It's available to Pro, Pro+, Business, and Enterprise subscribers, though Business and Enterprise admins have to manually flip a policy switch to turn it on — it's off by default at the org level, which tells you Microsoft itself isn't fully confident in shipping a third-party frontier model on by default to paying enterprise customers. That "third-party" framing matters more than it looks. Grok 4.5 shipped July 8 as a joint Cursor/SpaceXAI release — the first model release since SpaceX's $60B Cursor acquisition, as this blog covered at the time, complete with Cursor's own disclosure that an earlier snapshot of the Cursor codebase had accidentally leaked into the model's training data. Copilot is now offering that same model, built and marketed by a rival IDE's parent company, inside its own competing integration. It's a strange arrangement: Microsoft buying model access from the SpaceX-owned entity behind Cursor, to plug a gap in Copilot's own lineup — and inheriting whatever benchmark-contamination questions came with it. Copilot's model picker now spans OpenAI's GPT-5.x family, Anthropic's Claude models, and xAI's Grok — a genuinely multi-vendor menu, which GitHub frames as choice. It's also, less charitably, evidence that Copilot doesn't have a frontier model of its own worth defaulting to, and has to keep shopping. ## The "someday we'll unify this" admission One day later, on Microsoft's fiscal Q4 earnings call, Nadella confirmed what outside observers have been able to infer from the product surface for months: Copilot Chat, GitHub Copilot, Copilot Cowork, and Autopilot are four separate things wearing one brand, and Microsoft is building a single app to merge them. His framing was that Copilot is evolving "from chat to Cowork to Autopilots," and that this quarter Microsoft is bringing those pieces — including code — into one interface. No firm ship date beyond "coming this year," no detail on how a developer's existing GitHub Copilot workflow maps onto whatever the merged app ends up looking like. Take that at face value for a second: Microsoft is telling investors, in July 2026, that its AI coding and agent products don't currently function as one product. That's not a minor internal reorg detail — it's an admission that using "Copilot" today means picking among at least four different surfaces with overlapping but not identical capability, context, and pricing, and hoping the one you picked does what you need. ## Why this keeps happening to Copilot specifically Compare the shape of this problem to Claude Code, which has been one coherent terminal-native CLI since launch — same tool whether you're running it interactively, headless in CI, as a subagent orchestrator, or through the SDK. There's no "Claude Code Cowork" and "Claude Code Autopilot" running as separate apps that Anthropic has to eventually promise to merge. The reason isn't that Anthropic got lucky — it's that Claude Code was designed around a single agentic loop from the start, rather than bolted together from an IDE autocomplete plugin (Copilot's actual origin), a separate chat product, a separate async-agent product, and now a separate model vendor relationship layered on top. Microsoft's unification promise is the right instinct, six-plus product-generations late. GitHub Copilot launched as inline completions in 2021. Copilot Chat, Copilot Workspace, Copilot Cowork, and Autopilot each arrived as their own thing, on their own timeline, integrated into whatever surface made sense at the time — VS Code, github.com, a standalone web app, Microsoft 365. Grok 4.5's arrival this week isn't really about Grok; it's the same additive pattern playing out one more time, a new capability stapled onto the pile rather than designed into a single system, while the promised unification remains a "coming this year" line on an earnings call. ## What to actually watch The real test isn't whether Grok 4.5 performs well inside Copilot — third-party benchmark reruns for it remain thin and this blog won't repeat xAI's own numbers uncritically. It's whether Microsoft's unified app, whenever it actually ships, produces something that behaves like one product with one context model and one agent loop, or whether it's a shared UI shell wrapped around the same four separate backends with their existing seams still showing underneath. Given that Copilot just added a fifth vendor's model to the pile the day before promising to simplify everything, the smart bet is on the shell. --- **Sources**: [GitHub Changelog — Grok 4.5 is now available in GitHub Copilot](https://github.blog/changelog/2026-07-28-grok-4-5-is-now-available-in-github-copilot/), [Digit — Microsoft is building an AI super app, Satya Nadella says launch is coming soon](https://www.digit.in/news/general/microsoft-is-building-an-ai-super-app-satya-nadella-says-launch-is-coming-soon.html), [News9live — Microsoft posts record $331 billion revenue as Nadella teases AI super app](https://www.news9live.com/technology/tech-news/microsoft-posts-record-331-billion-revenue-satya-nadella-teases-ai-super-app-2994777), [Gizmodo — Microsoft will soon release an AI super app](https://gizmodo.com/microsoft-will-soon-release-an-ai-super-app-2000792653), this blog's [SpaceX–Cursor acquisition coverage](/posts/spacex-cursor-acquisition-done-xai-coding-empire-2026/)