Cursor Startプラン、Grok 4.5固定medium・non-fast制限が明記にCursor updated its official plan description for the newly launched Cursor…

Cursor, the AI-assisted code editor built by Anysphere, has updated the official description of its recently launched Cursor Start plan to make explicit a restriction that had previously gone unstated. As reported on Qiita, the revised text now says directly that use of xAI's Grok 4.5 on the Start tier is fixed to "medium" quality and "non-fast" response speeds. The edit looks like a light copy change, but it is the first time the company has openly documented the performance limits that Start subscribers encounter with this particular model.

The Cursor Start plan began rolling out on July 28, 2026 as a lower-cost entry point in Cursor's subscription lineup, and the wording update appeared shortly afterward. Although the change amounts to only a few words, it clarifies two separate constraints that materially affect how the model behaves.

The first constraint, "medium" quality, appears to describe the model's output or reasoning configuration rather than a different model entirely. In practice this suggests that Start users are served a mid-tier variant of Grok 4.5 rather than its most capable, highest-effort setting. The second constraint, "non-fast," concerns request scheduling. In Cursor's terminology, "fast" requests are prioritized and returned with minimal delay, whereas non-fast requests can be queued and served more slowly, particularly during periods of heavy demand. Locking Start to non-fast responses means users on that tier are likely to experience longer waits than those on higher plans.

Taken together, the two limits define a clear performance ceiling. Previously, the plan description did not spell out these boundaries, which left room for users to assume they were accessing the full-strength version of Grok 4.5. The updated language removes that ambiguity, allowing subscribers to judge whether the Start tier meets their needs or whether a higher plan is warranted for latency-sensitive or complex work.

For context, Cursor is a fork of Visual Studio Code that layers AI features such as inline completion, chat, and agentic editing on top of the familiar editor. It routes requests to a range of frontier models and, over time, has offered access to systems from providers including OpenAI, Anthropic, Google, and xAI. Grok 4.5 is xAI's model, and its availability inside Cursor reflects the broader trend of coding tools supporting multiple model providers so users can pick the one that fits a given task.

The distinction between fast and non-fast requests is a long-standing part of Cursor's pricing structure. Historically, higher tiers have granted larger allowances of prioritized requests, while lower or free tiers rely more heavily on queued, slower processing. Framing model access in terms of both a quality tier and a speed tier is consistent with how Cursor has managed compute costs across its plans, and the Start update simply makes that framework explicit for a specific model.

The move also fits a wider pattern across the AI coding market, where vendors such as GitHub Copilot and Windsurf differentiate plans by model access, request limits, and response priority. As inference costs remain significant, providers increasingly gate premium capabilities behind