Grok’s latest headlines span new tools, model performance, creative experiments, and questions about how AI handles sensitive topics. For readers tracking tech, the changes matter because they affect how people work, create, and get information. They also raise questions about safety and trust.
xAI, the company behind Grok, was founded by Elon Musk. Its AI assistant and model family now reach users through apps, paid plans, and developer tools. Reinforcement learning is a training stage in which a model improves through feedback on its outputs.
This update gives US readers a clear way to assess product changes, performance claims, access, cost, and public scrutiny. The goal is to help readers see what each announcement means for their own use, without treating a single test as the whole story.
Key Takeaways
- Recent releases include new voice, security, and sales tools.
- The Odyssey project has a seven-minute first episode.
- Benchmark results change with the task and test.
- Election testing used forced-choice API prompts.
- App, subscription, and API use have different costs.
- Public scrutiny includes nonconsensual sexualized images.
Readers can judge each update more clearly by separating product announcements from independent performance measures and public concerns.
What’s happening in Grok AI News right now?
Grok’s newest headlines range from workplace tools to election-bias findings and image-safety investigations. The election results came from a limited API test, not a test of every consumer conversation.
Grok Bot reached 418,000 weekly users on September 14, up 24% from the prior week. The figures point to growing use of workplace tools, though they cover one time period.
Enterprise releases and the Odyssey episode
xAI released Grok Voice 3.0 on October 14, 2026. On October 20, Grok Bot received an incident-detection and response update.
The SpaceX Super Intelligence Partner Network launched October 22. On October 27, xAI announced Grok Bot as a sales teammate for company work.
The Grok Odyssey episode is a seven-minute first release, made with Wonder Studios and director Dev Walker. Elon Musk had promised a full-length Odyssey film by the end of 2026, so the online episode is not that movie.
Odyssey uses generated video to tell a story, a format that raises questions about control and creative work. Readers comparing image tools can review this AI image generator guide for how creation and editing features differ.
Election bias and image-safety scrutiny
A forced-choice API study spanning nearly 500 races found Grok recommended a Republican candidate in 92% of answers. GPT-6.1 Sol recommended a Democrat in 97%, and agreement between the models stayed below 10%.
Forced-choice API results show how a model answered one test prompt, not how every user will experience it.
That distinction matters because only 8% of Americans say they use chatbots for election information. The test does not predict how the consumer app will respond in open-ended chats.
UK and EU investigations have also focused on nonconsensual sexualized deepfake images made with Grok. The Human Rights Research account of AI-made intimate images describes the safety and personal harms at stake.
Treat a model’s forced-choice API response as a limited test result, not as a substitute for checking candidates’ positions directly.
What does Grok 4.7 offer, and how does it compare?
Grok 4.7 pairs a longer training run for extended tasks with results that vary by benchmark. It scores higher than one rival on a coding test, but lower on the Intelligence Index.
xAI released Grok 4.7 on September 21, 2026. It used a larger base model than Grok 4.6 and a longer reinforcement-learning run, with more weight on tasks lasting many hours.
Training and capabilities
The training aimed to help the model check its work, manage longer context, and create documents and presentations. Its context window was 500,000 tokens in the October 12 comparison.
Grok 4.7 scored 62.4% on LatchBio’s biosafety benchmark. On HackerBench v0.3, it allowed 3.3% of risky dual-use prompts through.
The results speak to different needs: a biosafety test measures one kind of risk, while coding and reasoning tests measure other skills. The Vals AI Grok 4.7 profile reports results from additional task tests.
Benchmark results in context
The Artificial Analysis Intelligence Index v4.3.2 compares models across a set of tasks. On October 12, Grok 4.7 scored 46.4 at its top reasoning setting, while GPT-6 Sol scored 47.6.
On Terminal-Bench 4.0, Grok 4.7 scored 38.0%, compared with 37.3% for GPT-5.6 Sol. Those scores are close, and they apply to that coding benchmark.
Artificial Analysis measured Grok 4.7 at $3.74 per Intelligence Index task. GPT-6.1 Sol cost $0.72 per task in the same comparison.
| Metric and test date | Grok 4.7 result | Comparator model | Comparator result |
|---|---|---|---|
| Intelligence Index v4.3.2, October 12, 2026 | 46.4 | GPT-6 Sol | 47.6 |
| Terminal-Bench 4.0, updated October 12, 2026 | 38.0% | GPT-5.6 Sol | 37.3% |
| Cost per Intelligence Index task | $3.74 | GPT-6.1 Sol | $0.72 |
For a wider view of how assistants fit different work, see this comparison of ChatGPT, Claude, and Gemini. Compare models on the task you need, since the strongest result in one benchmark does not establish an overall winner.
How can you access Grok, and what does it cost?
US readers can use Grok through mobile apps, developer tools, and subscription plans, but those routes do not share one billing pool. The X Subscription costs $0 to $200 per month, while API use is billed separately.
Grok is available as an Android and iOS app, through the Grok API, and in Cursor and Grok Build. It also works with third-party coding harnesses, model routers, and cloud platforms.
Apps, API, and coding tools
The US App Store lists the Grok AI app as free with in-app purchases and names X Corp. as its developer. Android users can get the app through its official listing; an xAI APK is not needed.
Use an official app listing rather than an unverified APK download. For developers, API charges start at $2 per million input tokens and $6 per million output tokens.
Output tokens cost more per million than input tokens, and total token use drives API charges. This helps explain why a long response or heavy API use can cost more than a brief request.
xAI joined SpaceX on February 2, 2026. As of March 31, Grok AI features had 117 million monthly active users, while SuperGrok had 1.9 million paying subscribers.
The US user counts are dated figures, not a live total. A guide to AI chatbots and their paid plans can help compare options by workflow.
US subscription plans and usage
The monthly US prices for The X Subscription are:
- Free: $0 per month.
- Plus: $8 per month.
- Premium: $30 per month, with four times Plus usage, Grok Bot, and the latest Grok models.
- Super: $100 per month, with 15 times Plus usage and priority access at peak times.
- Ultra: $200 per month, with 50 times Plus usage.
Paid tiers share a weekly usage pool across Grok and Cursor. API use has separate billing, so subscription access does not cover API charges.
Check which usage pool applies before choosing a paid plan, since app and API charges do not combine.
Before paying, choose the access route that matches the task and check its usage limits as well as its price.
Conclusion
Grok’s latest developments are most useful when judged by the task, the evidence, and the cost of access. Model results depend on the benchmark, while app, subscription, and API routes set different costs and limits.
Readers considering a paid plan can compare the current app, subscription, and API terms before committing. A short trial of the intended task can show whether the tool fits a real workflow.
Check the current terms for the route you plan to use before starting a paid plan or API integration.
FAQ
What is happening to Grok AI right now?
Grok Bot was announced as a sales teammate on October 27, 2026, after several recent product updates. Grok Voice 3.0 launched earlier that month, and Grok Bot gained incident-response tools. The Grok Odyssey project also released a seven-minute first episode with Wonder Studios and director Dev Walker. The episode is not the full-length film Elon Musk had promised by year’s end.
Is Grok better than ChatGPT?
Grok 4.7 scored 38.0% on Terminal-Bench 4.0, ahead of GPT-5.6 Sol’s 37.3%. On the Intelligence Index comparison, GPT-6 Sol scored 47.6, above Grok 4.7’s 46.4. Those results point in different directions because each benchmark tests different tasks. They do not establish that one assistant is better for every user or kind of work.
Does SpaceX own Grok AI?
Grok is xAI’s model family, and xAI joined SpaceX on February 2, 2026. That company relationship does not by itself establish that SpaceX owns Grok. The name Grok refers to xAI’s AI assistant and models, while the partnership announcement concerned xAI joining SpaceX. Readers should distinguish that stated relationship from claims about ownership.
How many people are using Grok AI right now?
Grok AI features had 117 million monthly active users as of March 31, 2026. SuperGrok had 1.9 million paying subscribers at that time. These are different measures: the first counts monthly users of Grok features, while the second counts paying subscribers to SuperGrok. Neither figure is a live user total, and they describe different kinds of access.
Is there any AI better than Grok?
GPT-6 Sol scored 47.6, compared with Grok 4.7’s 46.4, on the October 12, 2026 Intelligence Index comparison. That result makes GPT-6 Sol the higher scorer on this benchmark alone. It does not show that GPT-6 Sol performs better on every task, or that Grok is a poorer fit for a particular user’s needs.
How much does Grok cost per month?
The X Subscription ranges from Free at $0 to Ultra at $200 per month in the United States. Its other paid tiers are Plus at $8, Premium at $30, and Super at $100 per month. API use is billed separately, based on tokens sent and received. A subscription tier and API access therefore have different costs.
Is Grok cheaper than ChatGPT?
On the Intelligence Index task-cost comparison, Grok 4.7 cost $3.74 per task, while GPT-6.1 Sol cost $0.72. Those figures compare the cost of completing benchmark tasks, not monthly subscriptions. The task costs also do not show what a particular user will spend across different workloads. Monthly plan prices need a separate comparison.
Is there a Grok free version?
Yes, The X Subscription has a Free tier priced at $0 per month. The US App Store also lists the Grok iOS app as free with in-app purchases. Free access does not mean every feature or level of use is included. Users who need more usage can compare paid tiers, while developers can use the separately billed API.
