Grok 4.7, Grok Bot and Team Bots: what's new, what it costs, and how it did in our test
Grok 4.7 costs the same as Grok 4.6. We tested both on the same coding tasks and read xAI's docs on Grok Bot, Team Bots and where your data goes.
xAI, which now calls itself SpaceXAI, released Grok 4.7 on 21 September 2026 and describes it as its "most powerful model for coding and knowledge work". A week later it added Team Bots, shared AI assistants for companies. We read xAI's documentation and tested Grok 4.7 against its predecessor, Grok 4.6, which costs exactly the same per token. Both models passed the same hidden tests. Grok 4.7 used about twice as many output tokens to get there, so at the same token price it cost us about twice as much.
What Grok 4.7 is
According to xAI's release notes, Grok 4.7 has been available on the xAI API as grok-4.7 since 21 September 2026. The announcement calls it "our most capable model for coding and knowledge work" and says it "works longer on difficult tasks" and "checks its own work more carefully". xAI writes that it uses "a new, larger base model compared to Grok 4.6".
xAI's model page lists a context window of 500,000 tokens, text and image input with text output, a knowledge cutoff of May 2026 and four reasoning levels: low, medium, high (the default) and xhigh. It also says there is "no text output limit". It is available through the xAI API, in Cursor and xAI's own coding tool Grok Build, and through model gateways such as OpenRouter, which is how we tested it.
What it costs
The price list gives two price levels, depending on how long your prompt is:
| Grok 4.7, per million tokens | Input | Cached input | Output |
|---|---|---|---|
| Prompts under 200,000 tokens | $2.00 | $0.50 | $6.00 |
| Prompts of 200,000 tokens or more | $4.00 | $1.00 | $12.00 |
That is exactly the price of Grok 4.6, as xAI's own announcement says: Grok 4.7 is "served at the same price and speed as Grok 4.6". A separate US regional endpoint, for requests that have to be processed in the United States, costs 10% more.
Compared with the other new models we covered recently, Grok 4.7 is cheap on output. GPT-6.1 Sol and Claude Sonnet 5.5 cost $2 input and $10 output, as we described in our Gemini 4 Argon article. Per token, Grok 4.7 is the cheapest of that group. As our test shows, the number of tokens a model needs matters just as much.
xAI's benchmarks: strong in places, not everywhere
xAI's announcement compares Grok 4.7 with Grok 4.6, OpenAI's GPT-5.6 Sol and Anthropic's Fable 5.1 on seven tests. These are xAI's own figures, which we could not reproduce. They are worth reading closely, because they are not all wins:
- Ahead: EEBench, an electrical engineering test (64.0% against 56.4% for Fable 5.1), and Harvey's legal benchmark (19.6% against 6.7%).
- Behind: on CursorBench 4.0, xAI's headline coding test, Grok 4.7 scores 46.3% and Fable 5.1 51.8%. On Terminal-Bench 4.0 the gap is larger: 37.6% against 57.9%. On HealthBench Professional, Grok 4.7 also trails both rivals.
The subtitle of the announcement says "Twice as fast, at half the price of comparable models", while the text itself says Grok 4.7 has the same speed as Grok 4.6. The speed claim seems to refer to the separate Fast variant, which we describe below.
Our test: Grok 4.7 against Grok 4.6
Because both models cost the same per token, the question is simple: does the newer model do better work, and what does it use to get there? We reused two tasks from our Codex test of GPT-6.1 Sol and GPT-6 Astra: fixing a bug in a function that reads durations such as "1h30m", and building a Dutch invoice calculation from a one-page specification. Each model got each task three times as a single request through OpenRouter, with default settings, and had to return the complete file. We then graded each answer with the same hidden tests as before.
| Grok 4.7 | Grok 4.6 | |
|---|---|---|
| Bug fix: hidden tests passed | 23/24, 23/24, 23/24 | 23/24, 23/24, 23/24 |
| Invoice spec: hidden tests passed | 15/15, 15/15, 15/15 | 15/15, 15/15, 15/15 |
| Output tokens, six runs | 39,966 | 19,376 |
| Cost, six runs | $0.246 | $0.119 |
| Time, six runs | 336 s | 250 s |
The scores were identical: every run of both models passed the same hidden tests. The difference was in how much each model wrote. Grok 4.7 produced about twice as many output tokens as Grok 4.6, almost all of them reasoning that you pay for but do not see: 35,858 reasoning tokens against 16,567 over six runs. At the same price per token, that made Grok 4.7 about twice as expensive for the same result, and it took about a third longer. It also varied more: its output per run ranged from about 3,400 to 9,400 tokens, Grok 4.6 from about 2,900 to 3,900.
In fairness to Grok 4.7, xAI says it "works longer on difficult tasks". Our tasks were not difficult enough to need that, so the extra thinking did not pay off here. On harder work it might. In absolute terms the amounts are small: all six Grok 4.7 runs together cost less than 25 cents.
As in the Codex test, the one bug-fix test that nobody passed checks a rule our task did not spell out, so we do not count it against either model. Two tasks and six runs per model is a small test: it shows how the two behave on clear, everyday coding jobs, not on long projects.
Grok 4.7 Fast: only in Cursor and Grok Build
xAI also offers Grok 4.7 Fast, which its documentation describes as "the same model served on faster infrastructure". It costs twice the standard token rates, or 1.5 times for long prompts.
The catch: it is "available only in Cursor and Grok Build", is not included in Grok Build's free tier and "is not available on the public xAI API". So developers who call the API directly, or use a gateway such as OpenRouter, get the standard speed.
What Grok Bot is
Grok Bot is xAI's assistant app, launched on 11 August 2026. The documentation describes "Bots you can keep around: AI teammates with names, jobs, and context that compounds over time". Each Bot works on a persistent cloud computer with a browser, a file system and a terminal, so it can carry out tasks rather than only write about them. It is a separate product from the Grok chat app on grok.com.
Access is not sold separately. According to xAI, Grok Bot "is included with every paid individual Cursor plan and with the Cursor Teams plan". You can also link an individual SuperGrok, SuperGrok Plus or SuperGrok Heavy subscription, and Cursor's help page adds X Premium+. SuperGrok Lite does not include it. Usage resets weekly.
Two details stand out in xAI's own security documentation. Grok Bot runs on Cursor's infrastructure, and "Cursor manages model selection": there is no model picker, and xAI does not promise which model does the work. We did not test Grok Bot, because it requires one of those paid plans.
Team Bots: one Bot for a whole team
According to xAI's announcement of 28 September 2026, Team Bots are in public beta on Teams and Enterprise plans. In the documentation: "A Team Bot is one Bot that your whole team talks to." Its owner sets it up once, with the plugins, secrets, skills and files it needs, and every teammate can then chat with it in the Grok Bot app or in Slack.
xAI says each person's conversations with a Team Bot stay private, and that usage counts against each teammate's own allowance, not the owner's. A Team Bot can hold up to 25 secrets, such as passwords or API keys; audit logs are only available on Enterprise plans. That combination, a shared assistant holding a team's secrets on a cloud computer, is worth a careful look by whoever manages security in a company.
For readers in Europe: where your data goes
This is the part we would check first as a European user.
- Grok Bot: xAI's security page says "Grok Bot computers run in the United States today". It adds that Cursor's US-only data residency programme does not apply to Grok Bot by default.
- The API: the regional endpoints page says xAI "may route requests between regions for capacity or reliability, so the processing location is not guaranteed". There is a US endpoint, but we found no EU endpoint.
- Training: for the API, xAI's security FAQ says it "never trains on your API inputs or outputs without your explicit permission", and keeps requests for 30 days for auditing. Teams can switch on zero data retention.
xAI's release notes mention that Grok 4.5 became available in the EU in July 2026. We found no such entry for Grok 4.6 or 4.7. In the consumer Grok app, the plans page still mentioned Grok 4.6 when we looked on 5 October.
How we checked
On 5 October 2026 we read xAI's announcement, release notes, price list, the Grok 4.7 page, the Grok Bot overview, Team Bots and security pages, the regional endpoints and API security pages, and Cursor's Grok Bot plans page. The screenshots are our own captures of xAI's documentation, cropped only. xAI's news pages blocked our automated browser for screenshots. We read the Grok 4.7 announcement as page text and made a chart of its benchmark scores from that; the Team Bots announcement we could only read through a summarising tool, so for Team Bots we rely on the documentation wherever possible.
For the test we sent each task as one request through OpenRouter, three times per model, with default settings, and graded the returned file with hidden tests that we had checked against our own reference solutions. Time, token counts and cost come from our script and OpenRouter's response. The owner of this site paid for the requests. A first run of the test gave usable scores but a faulty time measurement in our script, so we fixed the script and ran the whole test again; this article uses only the second run.
We did not test Grok Bot, Team Bots or the Fast variant. None of the images is AI-generated. Prices, plans and availability can change at any time.