Grok 4.6 is live with a first-week 50% discount
Cursor's newest first-party flagship Grok 4.6 launched on 2026-08-12 at $2 input, $0.5 cache read and $6 output per 1M tokens, with 50% off during its launch week. The official pricing page labels it Cursor's current newest flagship, and the 50% is a limited-time deal that reverts to standard pricing after the first week (source).
That gives you two immediate takeaways: there is a new flagship in Cursor's model pool to choose from, and if you have heavy work this week, the discount halves that cost.
Pricing comparison after the first-week discount
| Model | Standard Input | Standard Output | Standard Cache Read | 50%-off Input/Output |
|---|---|---|---|---|
| Grok 4.6 | $2 | $6 | $0.5 | $1 / $3 |
| Grok 4.6 Fast | $4 | $12 | $1 | not marked 50% off |
| Grok 4.5 | $2 | $6 | $0.5 | none |
All prices are USD per 1M tokens. Note the 50% is only marked on the standard Grok 4.6 tier—the Fast tier ($4/$12) has no launch-week discount, so do not assume it is also half price.
Pricing comparison within Cursor's own pool
Against Cursor's own first-party lineup, Grok 4.6 is the flagship tier, not the only option. The official pricing page splits first-party models into three layers (source):
| Model | Input | Output | Cache Read | Position |
|---|---|---|---|---|
| Grok 4.6 | $2 | $6 | $0.5 | Newest flagship |
| Grok 4.5 | $2 | $6 | $0.5 | Previous flagship |
| Composer 2.5 | $0.5 | $2.5 | $0.2 | Budget workhorse |
| Auto | $1.25 (input + cache write) | $6 | $0.25 | Smart routing |
The pricing comparison conclusion is straightforward: Grok 4.6 and Grok 4.5 share identical standard rates, so with the same price there is no reason to stay on the old flagship—new tasks should go straight to 4.6. The real decision is Grok 4.6 versus Composer 2.5, a 4x price gap that should be driven by task difficulty, not by "which is newer."
Pricing comparison against rivals: worth switching?
Benchmarking Cursor Grok 4.6 against mainstream models, it sits mid-tier: far cheaper than Claude flagships and slightly pricier than Claude's low-cost tier. Using input/output list prices:
| Model | Input | Output |
|---|---|---|
| Cursor Grok 4.6 | $2 | $6 |
| Cursor Composer 2.5 | $0.5 | $2.5 |
| Claude Haiku 4.5 | $1 | $5 |
| Claude Sonnet 5 | $2 | $10 |
| Claude Opus 5 | $5 | $25 |
The pricing comparison shows Grok 4.6 matches Claude Sonnet 5 on input (both $2) but charges 40% less on output ($6 vs $10), and undercuts Opus 5 by a wide margin. If you already use Claude models inside Cursor, Grok 4.6 is a cheaper first-party alternative on the output side. Claude figures are list prices—confirm on each provider's pricing page.
Quota, API key and the official entry point
Grok 4.6 is not a standalone billing endpoint—it shares Cursor's subscription and API key tooling. Key points:
- A manual Grok 4.6 pick draws the Token Plan API pool; only Auto routing draws the Auto + Composer pool, and the two pools are separate.
- Generate a CURSOR_API_KEY in the Cursor Dashboard Integrations page and use it via the Cursor IDE, CLI, Cloud Agents or Cursor SDK—no separate top-up balance is required (source).
- The plan tier sets included quota: Pro includes a $20/month API pool, Pro+ includes $70, and Ultra includes $400, with on-demand continuing at the same token rates afterward (source).
- Annual billing is roughly 20% off for a fixed monthly budget; use Cursor Token Plan for a subscription or the Cursor model pool for pay-as-you-go.
Things to watch
-
The 50% discount is launch-week only. It runs from 2026-08-12 through the first week, then Grok 4.6 reverts to $2/$6—do not budget the promo price as the long-term rate.
-
The Fast tier has no first-week discount. Grok 4.6 Fast is $4/$12, double the standard tier; unless you specifically need low latency, the standard tier is the better value.
-
Teams add the Cursor Token Rate on third-party models. Team plans charge an extra $0.25/M for third-party routing, but Auto Cost and first-party Grok 4.6, Grok 4.5 and Composer 2.5 are exempt.
-
This is Cursor's first-party model, not xAI's Grok. Cursor Grok 4.6 runs on Cursor's tooling and billing and is separate from the xAI open platform's Grok—pricing and API keys differ.