Elon Musk's AI company debuts a new flagship model with a DeepSearch feature. It's available first to X Premium+ subscribers.
Elon Musk's xAI has released Grok 3, its latest flagship AI model.
According to TechCrunch, the new model features enhanced capabilities and what the company claims is 10x more computing power than its predecessor. It also includes multiple models, with early access available to X Premium+ subscribers ($50/month).
Hilariously, Grok 3 also has a feature called DeepSearch — an agentic research tool that analyzes internet and X content.
This should sound familiar. Its xAI's answer to OpenAI's headline-stealing Deep Research agent, and it comes just days after Perplexity launched its own Deep Research product. So the frontier AI labs are all making it very clear that they intend to capture the second-mover advantage at OpenAI's expense.
Grok 3 & Grok 3 mini: Enhanced reasoning for problem-solving
DeepSearch: Research tool analyzing internet and X content
“Think” & “Big Brain” modes: Handles complex queries
Voice mode: Launching within a week
Enterprise API: Expected in the next few weeks
"Grok 3 is an order of magnitude more capable than Grok 2," Musk stated during the launch presentation, describing it as a "maximally truth-seeking AI."
Grok 3 was trained using approximately 200,000 GPUs in xAI's Memphis data center. The company claims the model outperforms OpenAI's GPT-4o on several benchmarks, including AIME (mathematics) and GPQA (PhD-level science problems).
The company announced plans to open-source Grok 2 once Grok 3 is "mature and stable," which they estimate will take a few months.
A new SuperGrok subscription tier is also planned, reportedly priced at $30 monthly or $300 annually.
In my opinion, the strongest case that can be made for using xAI over the tools from other AI labs is its integration with X itself. Because, as it happens, X is where most indie hackers hang out.
Use DeepSearch for “Social Listening”: Not only can DeepSearch’s ability to analyze X content help you do manual market research, but it will soon be able to power advanced market research applications. On a related note…
Early API Access: Sign up for the waitlist for xAI’s enterprise API to leverage Grok 3’s enhanced reasoning capabilities. Explore opportunities to bridge the gap between the upcoming open-source release of Grok 2 and Grok 3.
As with any new release, it's not all sunshine and rainbows. I'm taking the following factors into consideration while I try out Grok 3, and I recommend you do too:
Unclear API Pricing: Enterprise API costs haven't been disclosed yet, so tread carefully before investing a bunch of time and energy into new products that use the new model.
X Platform Risk: Rely too heavily on X and you might get burned. In this case, you need to consider not just the unpredictability of the platform itself, but of the platform's executive.
junk, don't support this bozo
Too many LLMs, I can this time to try them one by one.
That's actually great. More competition means faster evolution of features and reduction of prices. So good job xAI.
Personally, I find it hard to find a reason to use Grok and would like to get your thoughts on this. I use Claude for code generation with Cursor and use ChatGPT and DeepSeek almost sequentially to find answers to daily questions. As a Twitter extension, I think it doesn't work as a good summarization tool, since Tweets already have a length limit, when Grok summarizes a post it always makes the post longer 😄
Do you have any use cases for Grok in your daily routine?
Anyone tested the Grok 3 and it's free or paid?
Great breakdown, Channing! Grok 3’s promise of enhanced reasoning and features like DeepSearch is definitely intriguing—especially for those of us keeping an eye on advanced research tools. The integration with X could open up some neat opportunities for indie hackers in market research and social listening. That said, the unclear API pricing and the reliance on X are valid concerns. I’m looking forward to seeing if Grok 3 lives up to its ‘maximally truth-seeking’ claim and how it stacks up against competitors like GPT-4o. Thanks for the detailed overview!
Grok-3 claiming to outperform GPT-4o is a big deal. Has anyone tested it yet? How does it compare in real-world use?
Waiting for Grok 3 APIs
I,m quite greatful to have been introduced to a new AI chatbot that excels in coding, world knowledge and mathematics. Atleast now we can absorb knowledge mainly, focused on key contents intended for excellence, application of knowledge and innovation.
Exciting developments with xAI's Grok 3! The enhanced capabilities and DeepSearch feature sound promising, especially for X Premium+ subscribers. It'll be interesting to see how it compares to other AI models in real-world applications.
Great! I will try it. I hope it will be helpful for my website nohighprice, that I am building, for tasks such as generating blog articles and extracting coupons.
Well, if the model is trained by and on data from X. I have high hopes for it in the direction of sentiment responses from Grok 3.
As a forum visitor, I appreciate the insight on API pricing and potential risks. It's crucial to evaluate these factors before diving into new products.
https://www.youtube.com/shorts/OfWFjFeYtFU
Have you seen what Elon said about it? 😂
This comment was deleted 8 months ago